Revising the Airspace Model for commercial drone integration Krzysztof BOSAK
[email protected] v. 1.04 2016-12-18
Index 1.
My 2 cents .........................................................................................................................................................................................1
2.
Addressing shortcomings of the original proposal and discussing solutions...........................................................4
3.
4.
2.1.
Estimating the magnitude of the problem .................................................................................................................4
2.2.
Lack of definition of ground reference .......................................................................................................................7
2.3.
Defining emergency buffer above normal transit altitude band ......................................................................8
2.4.
Attribution of very low altitude band between 200…400ft to transiting sUAS .........................................8
2.5.
No provision for minimizing inter-drone collision while reserving broad altitude band ...................11
2.6.
Ignoring existing surveying and mapping drone applications.......................................................................12
Revising the Airspace Model for commercial drone integration ...............................................................................15 3.1.
Defining USGL (Unmanned System Ground Level) ............................................................................................16
3.2.
Defining PSLD (Passively Safe Light Drone) ..........................................................................................................21
3.3.
Proposing improvements on FAA obstacle reporting .........................................................................................24
3.4.
Defining boundary between GU airspace assigned for light unmanned transit......................................28
Legal status, patent issues and future development........................................................................................................32
References ..................................................................................................................................................................................................52
Document history 2016-10-01 v. 0.0
Beginning of analysis
2016-11-11 v. 0.9
First draft
2016-11-20 v. 1.0
USGL rendering and brief outline published online/ DIYDrones
2016-11-24 v. 1.0
Submitted for publication on sUAS News
2016-11-27 v. 1.01
Reformatting, proofing
2016-12-03 v 1.02
Traffic density simulation
2016-12-13 v 1.03
ATC sectional charts, full DIYDrones publication
2016-12-18 v 1.04
External elevator concept, extended roof operations, generalized grids into fractional dimensions
1. My 2 cents This document is an answer of Amazon Fulfillment Centre employee since September 2014 (Stower at Amazon FC WRO2) and reflects fully my best intentions of making drone delivery operational. I have solid knowledge in practical sUAS operations gathered during years of operation in European Airspace and observation of development of mentioned systems in US and elsewhere. My experience includes designing and flight testing from ground-up autopilot system since 2008 (at the time no open-source solutions based on modern IMU stabilization were available), having performed around 2000 logged and documented autonomous flights and similar number of non-logged flights, including operations at night, during rain, fog, snow, New Year date change, below freezing point, 30-mile VLOS and BLOS crosscountry flights (mostly roadworks construction mapping, but also guided delivery contest), high altitude meteo probing flights, stratospheric airdrop, surveying flight of active airport in Poland (2013) in cooperation with airport authorities (within CTR airspace) and several others, all using my avionics and UAV designs. The document under analysis is Revising the Airspace Model for the Safe Integration of sUAS [REV] made public by Amazon Prime Air team on July 2015 and widely available on the web. My revision of the original proposal discusses shortcomings, proposes coherent definition of (hopefully) publicly acceptable and mutually beneficial solution, finished by legal status of presented ideas and obvious opportunities for future developments. I believe it is in best interest of people to share sUAS/drone operational experience and address possible shortcomings as early as possible. We all want drone deliveries to be made safe and beneficial for societies. The worst possible scenario would be the idea being approached but then partially misguided, resulting in creating public perception that the costs and risks outweigh benefits. I am acting after observing total lack of answer from Amazon Prime Air recruitment to my applications for several positions in March 2016 answering to all published positions (around 10, including research positions and Technical Program Manager, a few of them to date remain in unprocessed state others were withdrawn without notification), being unable to conduct drone research since 2014 due to salary less than 4USD/hour, unable earn money allowing to obtain US Visa while working for less than 200-900 years (depending on estimations), after decision of local Amazon HR management that no drones will be built locally. I have decided to contribute to the future of aerial traffic with my 2 cents worth of discussion. It is important to understand that starting my career at Amazon at the moment of creation of its Polish subsidiary in September 2014, I see no further methods to discuss the issues with community other than via public channels, Other factors include availability of intellectual byproducts in October 2016, artificially maintained by global companies by depriving the local market that extend beyond simple physical works (even for persons with history of scientific work for world’s largest nuclear reactor designs), practical closure of US job market for Polish people willing to work legally on projects longer than 3 months and my feeling that ideas proposed by Prime Air Team are not even converging to the ideas implemented and commercialized by me around 3 years ago, when further development has stopped. This document can be also seen as merciful gift for a society valuing work of persons submitting obvious patent applications 20-120 times more than salary of a person who has developed most of patented devices several years before patent submissions.
Revising the Airspace Model for commercial drone integration
Page 1
On more positive side, concepts presented in this document are now available. We can now estimate social costs of not establishing UAV research centers in Poland, forcing local intellectuals to resign from repairing their teeth, buying a car or hiring a gardener. Let’s assume that at some distant moment in the future, drone deliveries will contribute 0.1% to Amazon company revenues. From the moment of creation of low-tech jobs in Poland in September 2014, more than 2 years have passed. Whenever drones will be introduced, delaying this research by 2 years is delaying business implementation by 2 years. Since at the moment of writing, based on patents which merely describe research ideas that are being analyzed, I can honestly say that my personal research dating 2009-2013 was simply more advanced and more intensive that observed activity of entire Prime Air team. Let’s suppose 5% of revenues are later or sooner being reintroduced via public taxes into US public services using various complicated formulas. In order to stay moderately humble for no scientific reason at all, let’s assume that contribution of the fact I was forced to work on regular worker position was contributing only 10% to this delay. From the above, taking Amazon revenues as 30 billion USD quarterly [AMZNREV], 120 billion USD yearly, I am estimating US social services losses because of misguided human resources
management
to
be
2 years x 120e9 USD x 0.1% drone contribution
x 5% lost of revenues reintroduced via taxes x 10% personal contribution = 1’200’000USD.
At least this
amount of money is lacking somewhere in US social system and I consider this to be economically justified feedback for forcing me to accept non-negotiable salary below poverty levels. But there are opportunities. Obviously one should respect independent solution of foreign companies or states to delay their research while limiting inflow of workers from certain countries. In order to use global resources properly, remembering that logistic business has very thin profit margin leaving little space for errors, one might consider moving drone research from regions that are technologically advanced, but very costly in terms of manual labor costs what makes small series prototyping and experimental integration with supply chain inefficient. One might consider moving away from areas where knowledge and wealth is high, but so is copying potential and land is mostly privately-owned, habitants are conservative, or flying sites are surrounded by mountains, forests and water what diminishes chance of retrieving lost drone hoping at least for feedback about reasons of failure at early prototyping stage. One might consider moving from areas where rainfall is so common that it is hard to maintain early prototypes flying in the field within visual range with high intensity, and move drone research from countries without Amazon logistics centers, which are neither in EU nor US as this creates cognitive dystopia among researchers and causes potential future technology export issues. One might consider selecting countries where weather conditions are supporting drone research business, allowing several months of different seasons equally distributed in order to represent flying conditions likely to be encountered everywhere. It would be wise to not deprive certain countries from research and management position that might be pivotal in supporting Continental European logistic operations in view of strict export rules and possibility of UK considering leaving EU. One would normally prefer performing prototyping in regions with limited population density close to cities, therefore with limited urban sprawl. For flight testing purposes one might prefer conducting research in areas where air traffic laws are representative for several countries, that is not almost exclusively ruled by military, where airspace has geometric properties allowing legal operations of experimental drone at altitudes which are already safe, non intrusive to environment and expected to be introduced in other countries as soon as their law matures, There are also human factors involved in understanding of innovation: had Wright brothers been concentrated on obstacle avoidance radars, nothing would fly reliably for much longer time, it ever. Nothing helps more than proper mission definition and focusing on objectives. Nothing is less important in robotic industry than
Revising the Airspace Model for commercial drone integration
Page 2
defining limits and conditions of human-machine interaction as the earliest, principal design factor for the entire system. My personal view on current issues is that somebody was pumping money earned every day by forcing me to do manual labor for years, while sponsoring exotic pseudo-research. Early adopters buying drones are not innovators, and up-scaling things is the most primitive way of achieving ill-defined goals, typical for soviet military that went bankrupt once upon a time with entire social structure altogether. Clearly, one should be able to extract mutually beneficial concepts even from potentially ambiguous situations. There is always time to learn and in order to maximize learning process I would like to invite Prime Air research team to work a few years in Poland, what might help convincing them about simpler ideas like automating CDROM storage shelves what is significantly simpler design task, but could be only achieved after research team has earned practical experience in handling those items. Other ideas include Amazon’s NonStop Drone Shop. Local society employing sometimes professional photographers would also welcome introduction or advanced digital photography technologies that emerged at the beginning of a new century, making our jobs more complete, more varied and hopefully paid above poverty levels. The author is no selling any products and all products mentioned that were produced and exporter by him were all discontinued in 2013 due to lack of research and investment funding. Hoping to invite discussion on all levels, I wish you all pleasant lecture focusing on scientific, social, manufacturing and legal aspects of commercial drone integration that can be done smoothly and safely with civilian technology available since 2000-2008.
Revising the Airspace Model for commercial drone integration
Page 3
2. Addressing shortcomings of the original proposal and discussing solutions 2.1. Estimating the magnitude of the problem As mentioned in proposal [REV], one can estimate 85000 manned aviation flights per day in US. By guessing average flight duration of 2h, we can estimate the number of airplanes in the air at any given time to be 85000/24h*2h=7083. It is claimed that this number will be dwarfed by commercial drone traffic creating misguided impression that extraordinary measures are needed; all this without digging into real-life numbers. If we think about all this in supply-demand paradigm, we will observe at first that those 85000 flights daily are supporting one of the most important industry branches in probably most air travel-oriented country in the world. Those planes are not flying for fun, but because somebody is paying for the tickets supporting entire industry and associated services. Let’s suppose commercial drone industry will achieve similar level of popularity and try to estimate resulting traffic volume. Estimating ticket price as rough guess of 300USD per air travel, we can assume all passengers of said flights will feel free to spend once at the end of their travel their money on a single drone delivery for missing equipment like personal hygiene items, being able to afford additional fee for delivery of goods while acting under high time pressure. According to USDoT we have about 900 million (9e8) passengers yearly [NPASS]. This is 2.465.753 passengers daily in US, leading us to guess identical number of drone flights daily to satisfy demands of those passengers with important motivation to move quickly while spending money. Assuming flight time 1h, we get 2.465.753/24h=102.739 drones in the air at any given time, or 2055 drones per US state. Let’s assume each state will host 4 drone delivery centers (about twice the current number of Amazon Fulfillment centers) what limits entire delivery drone population to 20-mile radius around them. This is pessimistic estimate suggesting highly localized concentration of drones in US airspace, resulting in 5026 sq. miles per state for 2055 drones at any given time, which is barely 0.41 0.41 drones per sq. mile at the peak of industry capabilities, capabilities, and only over area served by logistic centers. centers. Let’s assume that somebody has assigned all drones at equal altitude and tries to fly in suicidal manner exactly at this specific altitude, with airplane having a wingspan of 164ft (50m)=0.031 mile. With ‘suicidal trip length‘ of 40 miles over entire airspace over one of logistic centers, we are covering surface of 1.24 sq. miles, having only 50.9% chance to meet a single drone. Given the fact that this requires flying at altitude band at most +/-40ft (which is exactly altitude band an entire fleet of drones is able to hold without being bored), have we ever witnessed any event with manned airplane capable of holding such narrow band of altitude, except low altitude strike bomber with enabled active altitude hold autopilot and ground proximity radars. Then, we can spend more time (left as exercise for the reader) to figure out what is the chance of an airplane meeting a drone if a drone can actually pass above or under the wing of an airplane as high as entire altitude band allowed for drones, and then look for history books or air traffic records if there have even been a case of such ‘suicidal plane’ flying low over any area on such a long distance. My conclusion is that once e define a drone class not more dangerous than migrating bird, we can accept their presence in narrow altitude band practically anywhere in nonnon-controlled airspace, airspace, the larger area area is being served the better (for lower drone density). This means mentioned traffic would be completely safe in those conditions even if nobody would share his own position – i.e. completely non-cooperative traffic in case of major airspace management failure. In case of aircraft landing approach, we can assume descent rate of 10ft/s, which translates to 10s to cross drone-populated band (as if one would expect drones to invade airspace of a local airport). During this time, traveling at 200mph, an airplane covers around 0.55 miles,
Revising the Airspace Model for commercial drone integration
Page 4
sweeping with its wings a surface 0.017sq. miles, what translates to only 0.7% chance of meeting a single drone on approach if somebody mistakenly tries to land at place remote from the airport, crossing altitude bands used by drones at peak development of drone industry, assuming nobody makes evasive action and that drones is so large that in fact it intersects any vertical section of the aircraft (passing over wing is counted like a hit). All this keeping in mind that most spectacular crash landings like that of capt. Sullenberger involved simultaneous bird strike into both engines of several dozens of birds in order to disable the engines. When we are talking about birds, according to [NBIRDS] we have on average 15 billion (15e9) birds in US. Taking as rough guess that tries to underestimate number of birds in the air, we assume only 1% of their lives birds spend when flying. Dividing by US area, we can estimate 15e9/3’794’100 x 1%=39 birds per sq. mile at any given time, at any altitude they like, statistically everywhere including lakes, deserts, mountains and airports. If we exclude deserts and similar terrain, there will be even more birds per sq. mile. In other words, when one accepts flying aircrafts in non-cooperative bird environment accepting 39 non-cooperative birds per surface unit, one should accept easily 0.41 commercial drones per said surface unit in the most intensively drone-served area. Concerning ‘evil mastermind’ scenario, when we define a drone class with mass and collision properties corresponding to bird characteristics, for example 7lbs (3.17kg) we can obtain entire swarm weight per logistics center in the order of 513*7=3591lbs (1626kg), which poses negligible threat of any terrorist usage even if swarm control system would be naively designed by allowing strictly centralized command of drones flying simultaneously to one given location at exactly the same time. This makes swarming delivery drones very inefficient tool for terrorists, because even when hijacked, all of them would be lighter, more spread and softer object than randomly hijacked private airplane, and their commercial payload will be mostly absolutely harmless. If we define an unmanned object with kinetic properties of a migrating migrating bird which would be much likely to be encountered than a bird even in remote future of economic development, however flying at strictly narrow predetermined space without narrowing anybody’s capability of using called G Class airspace near ground level, we can create conditions that are never worse than current bird collision reality, while proposing additional rules which will make it completely negligible for practical traffic considerations. considerations. We will be able to offer significantly lowered collision probability and consequences merely by ensuring uniform density of drones using their predefined airspace, as a contrast to birds flocking behaviors. All the above can be achieved even without communicating exact drone position with anybody, what is an additional assurance in case of worst case incidents like total loss of positioning system or telemetry. Proposed idea is not ruling out extension of this altitude band for larger drone types that would need additional communication systems.
Revising the Airspace Model for commercial drone integration
Page 5
Drone density vs distance from logistic centre, 3...20 mile range, 500 drones airborne per area served, birds spend 1% of life flying
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Number of drones per square mile
45
Distance from logistic centre [miles]
40 35
Average drone density per sq. mile
30
Average bird density in the air over US
25 20 15 10 5 0
Assuming that any done can be sent to any range ordered from logistic centre between 3 and 20 miles, with 500 drones in the air all the time, we get the following drone density per square mile.
Drone density vs distance from logistic centre, 3...20 mile range, 500 drones airborne per area served, birds spend 1% of life flying
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20
Number of drones per square mile
100
10
Distance from logistic centre [miles]
Average drone density per sq. mile Average bird density in the air over US
1
0.1
0.01
Same as before plotted using logarithmic scale. Except for very narrow peak above logistic centre, overall density is uniform and totally negligible for citizens and even for most forms of air traffic.
Revising the Airspace Model for commercial drone integration
Page 6
2.2. Lack of definition of ground reference Authors draw segmentation of the airspace relative to ground, being totally oblivious what the definition of ground level is. If we keep Above Ground Level precisely tied to ground features at 4 ft resolution, the airspace becomes 3D jagged maze of non-navigable, narrow segments. There are also different definitions of ground level, for example general aviation pilot is assuming treetop level as ground level for planning his emergency approach. An airplane with radar or laser altimeter depending on circumstances and defoliation ratio measure either treetop or physical ground level. Ultrasound level sensor tends to underestimate altitude over grass and trees, optical sensor at best will consider building roof as ground level, and the list goes on.
Airspace model as suggested by Amazon Prime Air Team simulated above Bellevue, WA. Red band 400...500ft above ground for emergency traffic Blue band 200…400ft for high-speed drone transit, whatever means high-speed and everything below is local drone traffic.
The same airspace as above seen from drone perspective. Despite 200ft band height, flight path cannot be straight over typical drone delivery area increasing collision probability, power consumption and generating noise. No matter at what altitude we place altitude bands, defining them as placed Above Ground Level following all known definitions from geodesic science or any airborne distance measurement devices will result in jagged, non-navigable, unused vertical boundaries between airspace classes. As a solution, USGL geometric virtual ground level is being proposed.
Revising the Airspace Model for commercial drone integration
Page 7
2.3. Defining emergency buffer above normal transit altitude band Whatever goes up must go down, and it is unreasonable to assume any emergency will result in commercial drone climbing suddenly above its normal transit band just in order to stay there. Any light aerial vehicle that is suddenly crossing this upper altitude limit because of weather anomalies, could as well be sucked several miles up to the rainstorm cloud. It is obvious that any emergency stationary band should be below transit band. The idea of buffer can be resolved using two simultaneous approaches, by proposing strict USGL ground reference and by defining specific navigation scheme, limiting density of traffic in upper part of altitude band (see GU airspace band definition).
2.4. Attribution of very low altitude band between 200…400ft to transiting sUAS This idea is probably based on incomplete understanding of US air traffic law and FAA obstacle list management and minimal practical flying experience. Amazon Prime Air Airspace Design demands high-speed transit zone for sUAS between 200ft (61m) and 400ft (122m) Above Ground Level. The first observation is that one should demand for legal solution that is manageable. According to current US law regulations, Subpart C—Standards for Determining Obstructions to Air Navigation or Navigational
Aids or Facilities §77.17 Obstruction standards. (a) (1), national database of ground obstacles contains objects that are not anywhere close to airfields contains objects higher than 499AGL. It is true that according to §77.9 Construction or alteration requiring notice. If requested by the FAA, or if you propose any of
the following types of construction or alteration, you must file notice with the FAA of (a) Any construction or alteration that is more than 200 ft. AGL at its site. But, FAA is not obligated to classify such object as Obstruction to Air Navigation (according to §77.17), as this depends on geographic location of the object and its proximity to airports. As a consequence, updated on 56-day basis Digital Obstacle File emitted by FAA [DOF] which is mentioned ‘national database of ground obstacles’ may or may not contain objects 200ft…499ft high located somewhere in suburban area.. From technical standpoint, sUAS system can know his Pressure Altitude above virtual standardized sea level using pressure sensor, or its WGS84 or Mean Sea Level altitudes emitted by GPS, or its altitude above takeoff point using said pressure sensor without much technological effort (as implemented in most sUAS autopilots). One can reasonably implement local terrain elevation database inside sUAS system, knowing the offset between WGS84 altitude from GPS and local Ground Level. From those, current Above Ground Level altitude can be estimated. In absence of GPS (or more general any GNSS) signals, knowing its takeoff position and last known WGS84 altitude, a drone can continue updating its altitude using air pressure/barometric data with great precision, typically not exceeding 10m (40f) drift during hour-long trip [BARODRIFT]. This is compatible with 30-minute requirement of Amazon Prime Air delivery drones and provides reasonable backup even in case of GNSS signal loss. Using current drone technology we can hold narrow 40ft altitude band for 1h or more even when performing purely inertial navigation, after after GPS signal outage has occurred, in total absence of telemetry, using solely its recorded terrain elevation map and pressure altitude sensor. Unfortunately, structures like wind turbines tend to appear in quantities over vast fields in many countries undermining the idea of flying between them without knowing a priori their position. We should not expect any UAV system being able to detect a steel cable of moored radio mast that is twice as tall as actual flight altitude, while still being below altitude required by the law to be published; therefore a drone cannot detect such feature from close range with Forward Looking Camera cruising at 50mph and cannot avoid it. Even worse, avoiding wind turbine blade would require not only identifying the object, but also its motion. This is
Revising the Airspace Model for commercial drone integration
Page 8
above capabilities of multi-sensor flying creatures known as birds (there are several documented wind turbine bird strikes on youtube). It is therefore unreasonable to ask for sUAS to navigate during transit below altitude at which unpublished in national database objects might exist. It should be noted that so far the problem has been obscured by the fact, that any aircraft operator was practically banned of flying anywhere near such object:: 14 CFR 91.119 Minimum safe altitudes: General.
Except when necessary for takeoff or landing, no person may operate an aircraft below the following altitudes: (b) Over congested areas. Over any congested area of a city, town, or settlement, or over any open air assembly of persons, an altitude of 1,000 feet above the highest obstacle within a horizontal radius of 2,000 feet of the aircraft. So far there was always ample margin of 1000ft of altitude spacing above anything protruding form
the ground. This has allowed relaxed laws defining what an Obstacle to Air Navigation is. Taking a closer look at 53-WA.dat published August 15, 2016 [DOF] we can see at Washington state approx 5200 not dismantled obstacles, approx 1500 structures between 200 and 400ft AGL (Above Ground Level), 800 structures between 400 and 600ft AGL and 19 above 600ft AGL. However, looking at precision column as described in associated readme file, we can see that several structures have surprisingly low altitude measurement accuracy. For example entry 53-021645, catenary at Seattle, 47 30’ 50.00’N 122 17’15.00’’W is listed there because it is very close to the axis of nearby airstrip, being a recent entry updated in 2012. Only 40ft tall and apparently precisely located, has horizontal accuracy of class 5 (500ft) and vertical accuracy of class E (125ft). From the point of view of any automated navigation, this could be as well 40+125=165ft high. Since this is catenary, it is virtually blocking entire area along railway from automated drone transit using this database, if we take its location accuracy seriously (apart from the fact it is near to the airfield so it would be a no-fly zone for drones anyway).
Railway catenary 53-021645 as described in Digital Obstacle File. It is very precisely indicated (red structure, in the middle), but official precision for this object creates huge no-fly zone for any automated traffic in the area (yellow). Taking measurement precision seriously, the object could be located left or right to the airstrip, or even directly inside approach path. Another example, a pole at local stadium 53-021414, with listed height 80ft, but potentially as high as 205ft. This is already inside proposed altitude band for transiting drone. Lesson is that nobody ever tried to create a list of objects for safe, automated transit traffic at such low altitudes and it is unlikely such a list would be created ‘on demand’ or maintained with rigorous approach needed for small drones.
Revising the Airspace Model for commercial drone integration
Page 9
The only imaginable solution is to move sUAS transit altitude band well above 600ft (182m), because slicing any airspace below 400ft is pointless in view of existing FAA databases, collected during several years, and lower band limit of 200ft AGL is placing drones right at treetop level in undulating terrain.
Revising the Airspace Model for commercial drone integration
Page 10
2.5. No provision for minimizing inter-drone collision while reserving broad altitude band Proposed model is ignoring the possibility of using precise, thin altitude sub-bands to minimize (already very small) probability of collisions between drones. Attributing wide altitude band for high speed transit drones, that are supposed to be better equipped with sensors than localized amateur flights, is wasting ‘bandwidth’ of the airspace and has unnecessary collision potential. Original proposal allocates as much as 200ft band for High-Speed traffic without specifying why so much space is required for a sUAS system, which in practice can hold altitudes with 40ft (10m) precision and deviate no more than 10m horizontally from prescribed flight plan even during transitional GNSS outages. Manned aviation is already using this scheme: using Flight Level, expressed in hundreds of ft pressure altitude (those are ‘Rubber Feet’ units, changing their physical size a few percent every day with changing atmospheric pressure because those feets relate to ideal atmospheric pressure), one defines high-altitude air traffic corridors as interleaved bands, different for westerly or easterly traffic. This is a crude system built on fact that most civilian traffic on Earth travels east-west. The motivation is, with airplanes flying in one direction, one has: a) more time to spot b) more time to perform evasive maneuver within structural limitations of the airplane c) in case of sUAS, kinetic energy of eventual collision is greatly diminished, reducing the worst-case risks from disintegration of both drones in the air to slight structural damage that is handled by onboard systems leading to immediate mission abort and forced landing Proposed solution for drones is to define altitude band as a function of course over ground, LOLIMIT being lower limit of designated altitude band, HILIMIT set correspondingly.
COURSE ALTITUDE AGL = HILIMIT − ( HILIMIT − LOLIMIT ) ⋅ 360 The formula has the following consequences: * maximizes opposing traffic vertical distance * winged sUAS loitering (circling) above specific site must change its altitude constantly in order to match their altitude to course * hovering sUAS is already minimizing its collision energy against other objects, and could simply climb to specific band before accelerating to cruise speed * if we assume typical sUAS cruise speeds at 30-60mph a collision without course-dependent altitude could yield head-on collision kinetic energy
EK =
(m1 + m2 ) ⋅ (60 + 60)2 = (m1 + m2 ) ⋅14400 2
2
* a collision when both sUAS use course-dependent altitude, when the fastest drone crashes into the slowest cruising drone is
EK =
(m1 + m2 ) ⋅ (60 − 30)2 = (m1 + m2 ) ⋅ 900 2
2
(16 times less)
* a collision when both sUAS use course-dependent altitude, when the fastest drone crashes into a hovering drone is
EK =
(m1 + m2 ) ⋅ (60)2 = (m1 + m2 ) ⋅ 3600 2
2
(4 times less)
Proposed schema in order to work, must allow separation between opposing traffic not less than barometric altitude drift during proposed 60 min flight that is 10m/40ft. This requirement, assuming GPS/GNSS signal fails and the only means providing long-term course is magnetometer-based electronic compass, requires
Revising the Airspace Model for commercial drone integration
Page 11
spacing between 90deg and 270 deg course is at least 20m/65ft. Also, the change of required altitude during loitering will in fact change stationary orbit of winged sUAS to sloped ellipsoid elongated enough to allow sUAS to change altitude band in the process. The formula is putting sUAS flying north on the top of altitude band. While this originates from requirement to allow for circular loiter pattern of winged drones, the choice of preferred direction is intentional. On earth, statistically most present wind direction is from West to East, while air traffic also prefers east-west air lanes. It is well known that even loitering, a relatively light drone will spend most of its time in the part of the circle that is facing incoming wing direction and will potentially slow down after climbing depending on its engine power and control method. At the same time, low level civilian general aviations traffic, should it inadvertently cross altitude band reserved for sUAS, will be exposed to a subset of sUAS traffic that minimizes kinetic energy of those object relative to intruding manned airplane. Preferring northerly or southerly flying direction atop of drone altitude band in course vs altitude formula limits kinetic energy of loitering drone on top of the altitude band and increases sUAS traffic density inside designated altitude band.. All the above indicates that desirable altitude band width for sUAS transit altitudes is no more than 30m/100ft. 30m/100ft. This proposal has advantage of reducing typical kinetic kinetic energy of collision 16 times smaller than original document. document
2.6. Ignoring existing surveying and mapping drone applications Authors of mentioned articles claim in Airspace Design chapter, that airspace below 200 ft, will be reserved for non-transit operations such as surveying, videography and inspection. This is very worrying and is equivalent to halting such operations in civilian airspace. Most probably the misunderstanding of technical limitations of making digital, georeferenced maps used in 3D building site reconstruction and othophotomap generation led to this proposal. As discussed in ‘Secrets of UAV photomapping’ [SUPM], altitudes below 200m/656ft are rarely used for surveying. In the past, the limitation of US airspace has led to delayed development of this UAV application compared to EU. First of all, surveying operations are usually performed within Visual Line Of Sight (VLOS), however the mission endurance required for photomapping equals or exceeds flight duration proposed for goods delivery. Consequently, even large multicopters with endurance reaching 40min flight time are popular, but insufficient for covering entire building site, gravel pit, a dam or large motorway hub. Those flights are performed using winged platforms at altitudes around 200m(656ft)…300m(984ft) in European continental airspaces depending on requested ground pixel size ranging between 2.5..10 cm(1-4 in). Maximal allowed photomapping altitude of 400ft practiced so far in UK and US airspaces is barely sufficient for the task, requiring either using multicopter drones that have cruise speeds below 20mph, or restricting operations to almost steady weather that limits operations to a few days per month, making commercial application impractical. Next limitation relates to inability to process photographic material containing tall objects. Operational practice indicates that below 600ft it becomes impossible to properly align images with autumn trees, below 500ft it is almost impossible to join photos in the winter, at 400ft results of orthophotomapping uneven agricultural field with uniform visual patterns might be unpredictable. This is independent on actual processing method used, but is strictly related to large horizontal off-axis displacement of photographed objects at the edge of images. Additional problem arises with required processing power. Orthophotomap synthesis software for UAVs has in practice quadratic algorithmic complexity at its part selecting nearest neighbors, often ignoring geo-tagged images searching for second-chance best match. Lowering altitude form 650ft practiced in EU countries to merely 200ft, is increasing amount of information to be processed by (650/200)=3.5 times, while computation time raises between 3.5 and 10 times. For 1 square mile map processed using professionally georeferenced points, with demanding resolution of 1 inch pixel size, raises
Revising the Airspace Model for commercial drone integration
Page 12
computation time from one day to two weeks, using best available standalone PC and decreases resulting map equality. One must understand that several gigabytes of photos cannot be easily uploaded to cloud services; therefore one has no option to process such maps in continuous manner. All this keeping in mind, that quality of orthophotos gathered below 600ft AGL is inevitably inferior, and in case of build-up areas completely unsuitable for practical operations. Additional problem relates to the fact of rapidly increasing flight times required to cover photographed area in case of several, narrow photogrammetric lines required at too low altitude, raising flight times for areas 1-2 sq. miles which are normally perfectly within VLOS above 2h. This is technically almost achievable by drones weighting around 11lbs(5kg) commercially available since around 2010; however during 2-3h flight the problem of changing sunlight direction is changing color balance, making certain applications like infrared, NDVI and other agricultural research impossible to be applied over large fields within VLOS.
Low-altitude maps suffer from distortion making alignment of several photos over tall buildings impossible (left) As long as flight altitude is reasonably high (right) and so is image overlap, (here 650ft(200m) AGL 80% along and 60% side overlap), image processing software is able to select least-distorted objects during mosaicking [SUPM] Map area vs number of legs, 1h flight, 60% overlap or 2h flight, 80% overlap
Map surface km2
80 70
100m AGL
60
122m AGL 140m AGL
50
280m AGL 40
420m AGL
30
560m AGL
20 10 0 1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Number of legs
Map surface obtained with a drone is diminishing rapidly with flight altitude below 980ft(300m) AGL [SUPM] The best altitude band for surveying applications appears to be between 600ft 600ft and 1000ft, traditionally within Visual Line of Sight except for surveillance of long objects like road construction and power lines monitoring. As a rule, one should not make impossible using civilian drones for one socially beneficial application that is already implemented worldwide (photomapping, surveying), surveying), just to sketch a dream of another, potentially beneficial
Revising the Airspace Model for commercial drone integration
Page 13
operation (goods delivery). Obviously such relatively high altitude (from US and UK laws perspective), while absolutely necessary today and in the future for professional drone surveying, is contradicting with the idea of No-Fly zone proposed by the authors between 400 and 500ft or with local videographic band being placed under 200ft.
Revising the Airspace Model for commercial drone integration
Page 14
3. Revising the Airspace Model for commercial drone integration The goal is to redefine current airspace model that will: -allow safe operation of fully autonomous, lightweight commercial drones -minimize worst-case scenario to those that cause no permanent injury -increase current aviation safety by providing a new reliable ground level model for emergency landings -introduce Global Navigation Satellite Systems into all kinds of aerial navigation while circumventing blackout scenarios or other limitations -increase practical volume available for current General Aviation traffic by refining vertical airspace limit definitions that take into account aircraft dynamics -introduce surveying drone types that at current state of laws have significantly reduced operational potential compared in US and UK compared to the rest of world, particularly most EU countries Proposed actions include: 1.
Defining USGL, a new type of ground level reference for defining unmanned aerial traffic with possible extension as new generation ground proximity warning system for all air traffic applications
2.
Defining PSLD (Passively Safe Light Drone) that is capable of safe operation in non-cooperating environment using its most basic navigational properties
3.
Proposing improvements upon practices for documenting of man-made aerial obstacles in order to increase awareness allowing using spaces dangerous for large aircrafts by smaller drones and allow automated obstacle avoidance
4.
Proposing new airspace geometry for rural areas, integrating best existing drone usage practices, extending existing airspace for amateur aviation, mitigating the risk of inter-drone collision even in case of congested, totally non-cooperative non-supervised unmanned system traffic during GNSS and communication outage situations
Revising the Airspace Model for commercial drone integration
Page 15
3.1. Defining USGL (Unmanned System Ground Level) USGL is defined as imaginary surface described on regularly spaced grid in WGS84 coordinates used by current GPS receivers, having the following properties: 1.
USGL is defined for coordinates: latitudes 24…50 North, longitudes 125...60 West, all inclusive.
2.
Grid spacing is two arc seconds; values are defined for even arc seconds. The grid comprises (12560+1)*60*30=118800 grid points by (50-24+1)*60*30=48600 grid points.
3.
At each grid point difference between local ground level and WGS84 ellipsoid height is expressed in feet, which are written as signed integer values, spanning height -8000...32000ft. Values outside of those limits are reserved for future use.
4.
At certain points on both oceans Mean Sea Level is below WGS84 ellipsoid height, therefore USGL will be negative number at those locations
5.
At all places USGL is not lower than local terrain level or high tide water level, ignoring obstacles such as trees, buildings or other structures. It is never lower than Mean Sea Level.
6.
For any specific USGL grid point, any terrain elevation data used for constructing the grid closer to given USGL point than any other USGL point is not above USGL altitude at that grid point.
7.
USGL should be not lower than base of any object publicly defined as Obstacle to Air Navigation
8.
For all grid points, relative altitude slope shall not exceed prescribed 1:10 slope gradient (in all
higher than 200ft at date of its publication. directions, based on true distance between points). If a point placed above real ground level is so low that would require steeper climb or descent towards his neighbors, the lowest of a set of neighboring points point is artificially elevated until the condition is met. 9.
For any pair of grid points, a grid slope change by factor more than 10 is forbidden. In case of Vshaped valleys, the lowest point is raised until the condition is met.
10. USGL should be equal to airstrip elevation for all registered airstrips, unless it is impossible to maintain 1:10 slope gradient requirement which should be treated as priority condition. In such rare cases, USGL will be higher than airstrip level.
Proposed USGL coverage
Revising the Airspace Model for commercial drone integration
Page 16
A single USGL grid (yellow) cell projected on stadium in Bellevue, WA (2).
US GL Grid slope Less than 1:10 In all directions
vel d le oun r :100 g pe 1 5 True al slo in g ri O H~10ft
H~25ft
D~200ft
Basic property of USGL grid geometry is to be not steeper than 1:10 slope (8). USGL Lake/river
datum WGS84 al titude High tide
GPS MSL (Mean Sea Level)
Low tide
Relationship between USGL, GPS WGS84 and GPS MSL altitudes. MSL will not be used for definition of USGL (4) (5). e a x slop 1:10 m
USGL
1:10 max slope change Physical ground level
USGL over a canyon. Flying below is not forbidden but slope rule creates uniform navigational requirements in all directions (8) (9).
Grid spacing
USGL
Revising the Airspace Model for commercial drone integration
Page 17
Solid natural obstacles smaller than grid spacing affect neighboring grid elevations (6). This rule is important when source data is based on higher Terrain Elevation Model than USGL grid.
Grid spacing
Registered base height for navigational obstacle
USGL
Neighboring USGL grid points are not smaller than declared base height of obstacles for air navigation taller than 200ft USGL is not intended as automated landing aid, because it is not containing natural yet variable obstacles like trees, nor man-made obstacles Its purpose is solely for defining transit navigation and additional ground proximity warning reference for manned aviation. Vertical grid spacing (2) is approx 203ft (62m), horizontal spacing varies between 188ft (57m) south, to 136ft (41m) north. Total volume of data is around 11 GigaBytes, and is acceptable volume for selective internet download. 30mile long flight might correspond to 80x80mile area to be downloaded prior to the flight, which is manageable 20MegaBytes of data that covers all navigation situations for such flight. Proposed altitude range (3) (4) allows the same schema to be extended worldwide. A few exotic airfields have approach gradient steeper than 1:10, but USGL will be used as both real-time ground proximity warning for manned aviation as well as reference for defining transiting drone traffic, while acknowledging there are airports that offer no missed approach or are deep in valleys, therefore cannot be approached from all directions. For those locations USGL would be useless as ground proximity warning. Requirement of 1:10 slope ratio (8) will ensure symmetric properties for navigation over valleys, mountains, shores and rivers independently in direction of travel of cruising aircraft or drones. Mathematically it is limiting first derivative of elevation represented by USGL grid. The condition is very efficiently implemented by raising any given point if any of its neighbors is higher than certain threshold value added to current neighbor’s altitude. Requirement of limited rate of change of the slope itself (9) is flattening V-shaped valleys. Along with carefully selected specific grid spacing, this simple condition is creating a mesh adapted for dynamics of objects moving at speed 50mph in such a way that following maximum grid slope change they are not experiencing more than 1.12G positive acceleration (when any traffic participant in his own reference frame weights effectively 112% of its stationary mass). This means that any flying object navigating worst-case seesaw altitude band change, is loosing its energy during pull-up segments of flight that occur 50% of time (the others are pulldowns at 0.88G and require no additional lift nor extra lifting power). Resulting synthetic worst case energy loss for an object is not more than 12%*50%=6% for any object traveling at 50mph. In algebraic sense this condition is limiting second derivative of elevation. The condition can be efficiently implemented by raising any USGL point to minimum elevation calculated among its neighbors. Traditional airspace slicing is based on corridors and flight levels, headings and courses. It ignores knowledge of actual position for devising actual terrain altitude. This limits terrain detection to visual identification, radars, ultra-sound altimeters, optical flow sensors and radio-communicated pressures at airstrip elevation and barometric altitude sensors. Those sensors cannot provide enough accuracy to navigate over terrain at
Revising the Airspace Model for commercial drone integration
Page 18
altitude expected for automated, low weight drone traffic. Absence of GNSS/GPS signals will not be obstacle to determining actual USGL height, because drones knowing their kinetic properties and barometric sensor can navigate over virtual terrain for several minutes, being aware of changes of safe ground elevation while heading for nearby emergency landing site. It would be pointless to define USGL in terms of GPS MSL altitude. Mentioned Means Sea Level Altitude, while tempting as being intuitive and historically used for maps, is merely a polynomial fit atop of WGS84 altitude. This polynomial fit of difference MSL-WGS84 altitudes is based on 200km (WGS84), 100km(WGS84+EGM96) or 10km(WGS84+EGM2008) resolution gravimetric grid data, its actual implementation inside GPS modules is rarely specified and probably never verified by air traffic agencies, while definition of WGS84 altitude is inherently simple and well defined for all GPS receivers. A digital map can be published in interactive form allowing community annotations, allowing drone community feedback at all possible places, pointing our eventual difficulties and contributing to global map of safely navigable terrain level. Providing resources needed for hosting and implementing such services could be satisfied by commercial data sharing cluster services provider willing to pioneer in drone delivery market.
Ground level definition used for navigation should include terrain geometry and aircraft navigation capabilities. Typically, USGL definition will create virtual reference placed well above ground or sea level even over slightly undulating terrain (Bellevue, WA).
Revising the Airspace Model for commercial drone integration
Page 19
USGL rendered over Grand Canyon based on USGS 10-meter resolution Digital Elevation Model, (NED) covering entire continental US. Any airplane matching PSLD in climb angle capability is free to navigate in any direction on any surface parallel to USGL.
USGL experimentally rendered over isolated Ślęża mountain in S-W Poland with other airspace elements rendered in the background. Proper definition of ground level allows simple ground avoidance algorithms even for the most primitive robotic systems, offering a chance of negotiating ridges at any place for vehicles satisfying climb angle properties. The idea is applicable globally with limited data storage efficiency in Polar Regions.
Revising the Airspace Model for commercial drone integration
Page 20
3.2. Defining PSLD (Passively Safe Light Drone) PSLD is Small Unmanned Aerial System (a drone) having the following properties: 1.
All-up weight below 7lbs (3.17kg).
2.
All sizes below 4ft (1.22m).
3.
Cruise speed between 25mph (40km/h) and 50mph (80km/h).
4.
Is able to navigate in winds equal to half of its designed cruise speed.
5.
Is able to maintain powered flight climb slope relative to ground 1:10 also in case of tail wind equal to half of its designed cruise speed, at Density Altitudes [DALT] compatible with its designed operational conditions.
6.
Total kinetic energy during cruise speed including rotational kinetic energy stored in moving parts like propellers not larger than 1200J.
7.
Is using electric propulsion.
8.
Propellers or other rotating parts providing forward motion should be placed behind lifting surfaces, or at the rear of vehicle.
9.
All propellers including lifting propellers or other rotating parts providing lift should be screened with a net or other airframe structure so that the airframe can be stopped from reaching human body using human hand as protection, without contact of this hand with rotating part.
10. Is not carrying any flammable liquids or chemicals toxic to water environment, except those being hermetically sealed in battery 11. Is able to float up to for 7 days upon crashing on water 12. Contains independently powered light-weigh proximity location device with range of at least 50ft over open field, able to function for at least 7 days 13. Use of metals, except for: electronics, engine, battery and sensors is forbidden 14. Use of materials that create sharp edges upon breaking is forbidden for construction, including carbon fiber and glass fiber and glass 15. It should not contain objects protruding from the front, in direction of flight. Any surface of fuselage or other protruding object should be no less than 1x1in in diameter, including sensors. Propellers generating lift force be shielded from the side of direction of travel. 16. Construction materials used for airframe construction should not be denser than 45lb/ft3 (hard wood) 17. Should not short-circuit causing battery fire or leak when landing in salt water 18. Should be covered with non-conductive material and painting 19. Is using GNSS as guidance, with ability of dead-reckoning to nearest Emergency Site in case of signal loss or GNSS jamming. 20. Will contain memory of operator-provided automated recovery sites away from human settlements that are used as Emergency Sites. Quantity and spacing of those sites depend on system’s ability to navigate within prescribed altitude band. 21. Has ability to maintain altitude accuracy due to pressure sensor drift not less than 10m during its entire flight and is able to use this altitude as backup in case of GNSS signal loss. 22. Has no optional geometric or design configurations in which it could be reconfigured by the user as having properties exceeding defined drone class. 23. PSLD properties are declarative: operator using a PSLD class sUAS is declaring its conformity. One is held accountant for any damage or incidents due to exceeding or not meeting one or several mentioned parameters.
Revising the Airspace Model for commercial drone integration
Page 21
24. Each example of PSLD drone should contain identification plate mentioning ‘PSLD compliant’ along with personal ID, Driving License Number or company’s Tax ID allowing identification of the operator, along with contact phone number. 25. Mentioned contact and ID information should be printed, written or permanently affixed using waterproof and UV radiation proof technology in two examples, both on upper and lower surface of a drone at easily visible place, using font size not less than 16pt. Those properties are minimal and are intended of allowing a system that is safe, but will allow socially responsible integration of useful drones. Applications include documenting public infrastructure, road works, and goods delivery. Proposed construction limitations are based on operational experience, are compatible with several existing photogrammetric drone systems and might be reused for definition of more demanding sUAS subclasses. Total mass limitation (1) is reflecting mass of a chicken used to test aircraft engine ingestion and windscreen resistance [FARBI] – however kinetic energy of proposed drone class is several times smaller vs typical obstacle or slow flying general aviation aircraft. Requirement of total kinetic energy (6) is reflecting the fact that large amount of energy could be stored in moving parts, which we want to declare as unnecessary. Building elements like windows, balustrade, and roof parts are routinely certified to be resistant to 1200J or energy delivered in form of kinetic energy [BUILDRES]. 1200J is also corresponding to 150% of max energy for proposed cruise speed 50mph (3) for an object weighting 7lbs (1). Including rotational energy (6) in the definition is very important because each rotating rod (a good approximation of a propeller) has moment of inertia [INERT], For example 27-inch propeller APC LP27013 [APC] weighting 9.52oz rotating at 5000RPM has easily 1450J of rotational kinetic energy [RKE] stored in itself. Unless exotic technologies like Inertial Nullifier are invented, even patents like ‘Propeller safety for automated aerial vehicles’ [PROPSAFE] suggesting a method of cutting power to a propeller in automated aerial devices upon a detection of collision with an object (otherwise used in commercial product classified as ‘brushless ESC motors for UAVs’ around the world since more than a decade) would not prevent delivering kinetic energy stored in the propeller itself to an object. On the other hand any instant propeller braking mechanism would need to absorb as much energy as entire drone would have by itself traveling with maximum cruise speed. Requirement of minimal cruise speed (3) is ensuring that PSLD can navigate in reasonable time back to its landing site in case of average winds, ruling out most blimp designs or a few particularly slow multi-rotor designs. Requirement of buoyancy (11) is important for protecting public water reserves in case of worst-case scenario and loss of airworthiness and is linked with lightweight radio beacon requirement (12) that will both ensure minimal weight of such telemetry, and ability to recover drone from approximately known crash site. It is important that telemetry requirement is purposely limited performance to popular self-powered key-fob or pet search gadgets, which do not increase significantly drone weight and consequently limit total kinetic energy (6) of the flying object. Any requirement of more powerful telemetry leads to enforcing potential radio interference, increased power requirements or mass or flying objects that are likely to appear. Requirement of low density of construction materials (13) (14) (16) is a guarantee that PSLD will be chopped by general aviation propeller or decompose onto pieces upon contact with aircraft structure. Limiting the use of carbon fiber, metals and glass is limiting high speed debris and splinters.
Revising the Airspace Model for commercial drone integration
Page 22
Requirement of non-conductivity (18) along with maximum size/wingspan/length (2) is limiting potential for aerial power transmission line shortcuts according to [TLIN] which requires 1.5m (60in) spacing. Definition is not strict as given enough voltage almost anything is conductive, but obvious exclusion include metallic paint, ‘wired flying cages’, to some extent carbon fiber and solar panels. One must remember that having smaller wingspan or width than cable distance is not excluding arcing, therefore non-conductivity requirement must be kept. Declarative character of PSLD (24) is not different from European EC conformity declaration: a manufacturer declares that given device meets specification, sometimes using third-party certification if one wants additional cover when things go wrong specifically because the manufacturer was lying, but it is not imposed to let every prototype pass rigid governmental registration procedures. This approach guarantees reasonable prototyping of innovative ideas, compared to current aircraft registration laws which are lengthy process focused on enforcing iterative approach for otherwise very mature manned aviation industry.
Name
Year of introduction
AllAll-up weight (AUW)
Cruise speed
Payload
Return flight range
Total kinetic energy
EasyUAV photogrammetric [ESYU] discontinued
2010
3.3lbs (1.3kg)
31mph (50km/h)
0.44lbs (200g)
10mi (16km)
126J
Parrot Swinglet cam [SWING]
2011
1.55 lbs (0.7kg)
22mph (36km/h)
0.45lbs (195g)
5.6mi (9km)
35J
Trimble Gatewing X100 [GATE]
2012
4.8lbs (2.2kg)
50mph (80km/h)
0.44lbs (200g)
18mi (30km)
545J
PSLD virtual
2016
7lbs (3.17kg)
50mph (80km/h)
3lbs (1.36kg)
10mi (16km)
785J
Comparing PSLD concept with existing commercial drone systems. Those devices were not optimized for payload. With around 50% range reduction, their payload would increase to 30%…40% of AUW (up to 3lbs/1.36kg). Requirements concerning simple identification plate (25) are based on the following assumptions: - law-abiding operator might want to have his drone returned after unsuccessful loss in remote area - citizens might want to have a method of pursuing malicious drone users - government officials might want to reuse legally accepted method of identifying the owner apart from obscure, often missing or unavailable logs or telemetry data - in suspicious cases law officers could confirm operator’s identify and ID on the drone upon landing at any moment confirming that the user is matching the drone - privacy-conscientious operator might want to conserve his privacy and replace his name by officially acknowledged ID number, in order to avoid unjustified or hurried accusations - alternative solutions using custom-issued or license-based approach would only delay drone prototyping to the point of non-introduction
Revising the Airspace Model for commercial drone integration
Page 23
3.3. Proposing improvements on FAA obstacle reporting 1.
Increase precision of official obstacle list precision both in altitude and in position at least to class 1 A as defined in [DOF].
2.
Create uniform definition of Obstacle for Air Navigation as anything above 200ft, instead of progressive definition for objects between 200…500ft related to distance of official airports.
3.
Include additional information about object radius in feet imagined as a cylinder encompassing object at the base and at any height, for all obstacles for which height+declared altitude precision is equal or greater than 200ft.
4.
Specify obstacle for air navigation precision as 1/10 of arc second, expressed in WGS84, for all obstacles for which height+declared altitude precision is equal or greater than 200ft.
5.
Supplement Digital Obstacle file with unique ID for each state for each connected list of objects that
6.
Publish national layout of power transmission lines in a version restricted for air traffic management
are connected with cables or other structures. (masts, height, precision, positions assembled in lists of connected points). While management or obstacles for air navigation is proposed in very well defined digital format available online [DOF], it is unclear why precision of determination of altitude and position remains very liberal (1). Digital Obstacle File precision, WA 1000 900
Height above ground[ft]
800
Height+error band Height as listed
700 600 500 400 300 200 100 0 0
500
1000
1500
2000
2500
3000
3500
4000
Obstacle sorted by height
For example in Washington state there is significant portion of obstacles that are apparently precisely defined, not dismantled, can be found with laborious searching on public maps, yet precision of their height estimation is often worse than measured value. For comparison, Polish law since 2003 demands precision determination to be 3m (10ft) far form airport (what corresponds to FAA class B precision) and 0.5m (2ft) near airport (better than FAA class A precision). This is contrasting with US data and is creating a lot of confusion about validity of FAA obstacle list when processed as a whole for automated traffic purposes (1). The idea of uniform definition of obstacles for air navigation (2) is that while it increases total number of objects roughly three times, it also creates more rigorously defined airspace between 200ft and 600ft for general aviation traffic participants that in a few years time following development of automated obstacle detection might want to have access to reliable database of objects. It must be understood that as long as drones are to be included in the laws and treated as air traffic participants, any rural area will become an airport in virtual sense. Even today, from the point of view of occasional ultra-light airplane flyer taking off from local
Revising the Airspace Model for commercial drone integration
Page 24
grass, non-uniform definition of object height is already creating a potential of missing important data (like newly constructed wind farm) for leisure flight planning. Including object radius (3) is very important. This data is almost impossible to extract from existing databases, yet easy to be declared by object builders or maintainers. Object of particular interest include groups of masts, propeller diameter for wind farms, moored cable lines for radio towers. The intent is to assist creating nationally recognized database of objects as side effect of existing procedures used for declaring construction or modification of objects taller than 200ft. Precision requirement (4) is similar to Polish air laws, except that it is a little more strict as it requires 1/10 arc second horizontal position precision everywhere, not just near airfields. It is routinely practiced to obtain accuracy even higher than this requirement using ordinary survey grade GPS, therefore a measurement itself should not be a problem. Modern stationary geodesic GPS easily achieve (a few) inch accuracy which is already 10-50 times better than 1/10 arc second requirement. When we consider practical application of mapping road works construction using existing drone system, we would want to estimate area occupied by given obstacle with sufficient precision, knowing that we can perform this relatively easy drone mapping operation with a small automated drone. Using current rules, when we are calculating potentially minimal airspace occupation by any single radio or cellphone mast that is barely 200ft high, we can easily achieve accuracy like +/- 1 arc second and adding obstacle radius, say 10ft, to 1 arc second which is around 100ft, we can have entire road crossing obscured by virtually reserved airspace by a single radio mast. This requirement corresponds to asking for horizontal class position accuracy raised to or slightly above current FAA Class 1 (20ft) for all obstacles above 200ft. For wind farms or for any other object this is not an indication of possibility of safe navigation directly near given object, but a base for further specification in the laws about safe navigation margins for different automated air traffic participants. Proposed improvement once coupled with USGL reference, will clean-up airspace and provide ‘breathing space’ increasing decision time during emergencies for general aviation between proposed drone altitude band and mentioned obstacles, so that those completely unequipped for aerial drone communication (what means during prototyping of drone management traffic systems – strictly anybody) could stay at their formerly practiced local aviation routes used for leisure while operating below thin layer of practically nonexistent drone traffic – should one ever irrationally fear it. The idea is that while this band for all statistical collision calculations will be negligible, there will be still enough time for a pilot to completely clear his mind from drone traffic and concentrate on proper emergency landing, preferably coupled with USGL as emergency navigation aid.
Revising the Airspace Model for commercial drone integration
Page 25
Large airspace zones with structure collision potential reserved around Seattle, WA only because of seemingly random attribution of Obstacle for Air Navigation precision of reporting. Small red circles visible inside yellow zones are obstacles themselves, assuming 30ft object radius.
Seattle, WA near airstrip: power transmission lines are represented only by sparsely placed objects; completely ignoring any descriptions of cables that are attached to them. Those cables are completely undetectable with all practical aviation landing aid instruments, also making any automated traffic assistance incomplete. Improvement (5) suggests adding information to database that is being transmitted by FAA as constructed objects are made by a single operator, which is lot during processing by FAA. This is essential of understanding what obstacle really is. It is not enough to annotate or mark on traffic maps a list of isolated objects if they are spaced by long distances and interconnected by cables, each of them can be critical hazard for emergency helicopter planning approach, for a drone trying to deliver goods, or for large civilian airliner using virtual rendering of terrain below as last resort collision avoidance support in case of emergencies. A solution is to supplement list of obstacles with additional column with an index that simply enumerates groups of objects which are interconnected. Improvement (6) is related to the fact that among all different energy pipelines and other installation maps available on US Energy Information Administration, just the list of Electric Transmission Lines is not published [EIATL] due to copyright issues. Absence of this data for public audience is reducing flight planning safety for everybody, including general aviation, low altitude emergency aviators, inventors, researchers and commercial drones. Limited publication (geometry only) of information with free access rights is necessary for creation of automated flight avoidance databases. Actual situation appears to be that one can find all US border crossings of oil and gas pipelines, processing plants and border crossings of mentioned electric transmission lines for convenient sabotage planning, but cannot find collective information about their geometry trying to fly safely at low level above US. The same problem concerns high voltage Electric Power Lines.
Revising the Airspace Model for commercial drone integration
Page 26
Already at this coarse map scale, no data about Transmission Lines is available for general public in US (grayed option, lower right pane, EIA U.S. Energy Mapping System) even if remaining objects of strategic significance yet harmless to drone navigation are enumerated in detail. Bottom line is that US infrastructure is well documented from the point of view of automated air traffic, but neither aerial cable infrastructure nor aerial traffic are protected properly because the data is sometimes hidden from the public, and is not uniform across power operators and states. Even when this data is published in graphic or paper form, it is published using methods making it totally unsuitable for digital processing and flight planning.
Revising the Airspace Model for commercial drone integration
Page 27
3.4. Defining boundary between GU airspace assigned for light unmanned transit 1.
Define lower level or E class airspace as defined in feet above USGL, instead of AGL.
2.
Designate altitude sub-band for light unmanned air traffic between 600ft and 700ft above USGL as GU airspace .
3.
For all legal purposes GU airspace is G class airspace.
4.
All Beyond Line of Sight commercial drone flights must fit within GU band and be performed with drones meeting PSLD requirement.
5.
All unmanned transit within GU band must obey course to altitude decongestion formula:
COURSE ALTITUDEUSAGL = 680 ft − 80 ft ⋅ 360 6.
Drones must travel within designated bands at all times except for landing and takeoff purposes.
The only place imaginable for unmanned aerial traffic at its present state is to stay below controlled airspace area in order to not overload existing air traffic systems designed specifically for human-based interaction. Large parts of West Coast and almost entirety of US East Coast is almost exclusively covered with E class airspace starting above 700ft AGL. Redefining existing airspace geometry could incur huge costs.
Class E airspace above 1200ft, Class G below
Class E airspace above 700ft, Class G below
Example airspace sectional chart for Seattle area, WA. Unfortunately, it is impossible to define proper placement of vertical altitude boundary between air traffic participants between G and E class airspace by air traffic participants. Stated 700ft AGL altitude is already very close to the ground, but is not taking terrain relief into account as no globally applicable definition was ever provided. What is worse, an aircraft, a glider or paraplane trying to stay below 700ft limit in G class airspace must stay well below this altitude. Air Traffic Controller is not able to instruct E class airspace user (above 700ft) to fly at specific altitude with precision, as he could not guarantee that by flying at altitude equal to local airport elevation +700ft one would not fly so close to the ground that it would make impossible to not collide with G class airspace users. ATC could instruct anybody by radio about flying at specific altitude which is a multiple of hundreds of feet, even if only local radar would allow for this precision measurement. Therefore all what ATC controller could advise to E class airspace user is to stay for example at ATC+800ft, but even this can fall below 700ft distance from the ground in relatively flat coastal areas. A conclusion is that at present state both G and E class airspace users must stay within significant distance from vertical airspace boundary: G class user because he might be devoid of equipment required to enter E class airspace (including ADS-B transceivers for collision avoidance), while well-equipped E class airspace user cannot afford regularly
Revising the Airspace Model for commercial drone integration
Page 28
entering G class airspace below as one might encounter not easily detectable under-equipped manned amateur aircraft. Because of highly variable geometries depending on specific altitude definitions, this region of airspace is according to the laws simply underused. Current region around 700 ft altitude do not offer guaranteed position reporting in its upper part (E class) nor enough geometric freedom for the lower part (G class).
Green AGL700ft above ground level understood literally or measured with proximity sensors ATC+600ft, 700ft and 800ft (orange, pink, red) are altitudes in hundreds feet above nearest airport that could be communicated to a pilot by Air Traffic Controller FL7 (hi and lo, blue) are flight levels normally used for defining air traffic at high altitudes, corresponding to 700ft distance above sea level only during strictly defined International Standard Atmosphere conditions. Those conditions may shift geometric altitude easily 30ft up or down depending on current temperature and air pressure.
New definition of lower limit for E class airspace (or upper limit of GU class airspace) 700ft (violet) is measured above USGL. At the same time 600ft (blue) lower part of GU class airspace is in most cases above traditionally understood imprecise definitions of 700ft AGL related to altitude above remote airfield or abstract definitions related to altitude above sea level.
Revising the Airspace Model for commercial drone integration
Page 29
Proposed airspace model has the following advantages: 1.
Air traffic maps are unchanged.
2.
No legal redefinition of G or E class airspaces is necessary.
3.
The idea is not requesting any custom-build electronics except minor code changes for the autopilots.
4.
The idea is not incurring weight, geometric or significant cost penalties for drone designs, inviting wide acceptance.
5.
The idea uses previously unused airspace because of lack of practically verifiable vertical boundary
6.
E class-equipped aircrafts can use GPS receivers integrated in their ADS-B transceivers to provide
definition. data to ground level warning system. 7.
G class users are benefiting from database of ground level using GPS receivers, having more incentive to use also position broadcasting devices.
8.
G class upper limit shape becomes unrestricted concerning the direction of travel.
9.
Separation between different classes of aircrafts from E and G airspace is filled with harmless drones, allowing tighter use of airspace. Safety is increased between non-cooperating aircrafts because aircrafts equipped for different style of flying (transponder-equipped or not, visual VFR or instrumented IFR rules) are now separated not only by virtual line drawn by the laws, but also physically.
10. PSLD drones operate at altitudes where they are inaudible, can provide efficient survey grade maps while yielding reasonable resolution and surface yield per flight. 11. Simple non-collision rule is implemented for initially totally non-cooperative drone environment. 12. Drone traffic density is higher at lower bottom part of GU band, allowing for blind (instrument) flying in E Class airspace with more safety even along its lower boundary. 13. More space is provided for emergency landing for non-powered air traffic like gliders above terrain. 14. Traffic band is placed above region where currently inconsistent air obstacle reporting exists. 15. Navigation planning is symmetric in all directions, concerning climb ration and power requirements allowing simple fully automated navigation as well as conceptually simple remotely piloted drones. 16. The system can be extended upwards optionally providing for 1200ft USAGL E class airspace lower boundary where it exists above rural areas. 17. Since traffic requirements allow for safe decongestion in uncooperative inter-drone traffic, beyond line-of-sight flights are equally safe. 18. GU altitude band defines clearly evasive maneuver geometry by binding course with altitude. Drones trying to get away from each other will automatically also select non-colliding altitudes; therefore the idea will also benefit fully motion sensor equipped drones. 19. Cooperative navigation rules are set between drone classes that perform radically different missions, like photomapping, security and goods delivery. 20. No difference is requested or desirable between amateur traffic and commercial solutions, allowing synergy at prototyping level and introducing amateur drones into well defined flying patterns, increasing general air law awareness of amateur society. 21. Strict geometric definition provides legal ground for eventual incursions, 22. Allocated airspace is applicable over entire US territory except for controlled airspaces over airfields and other excluded area.
Revising the Airspace Model for commercial drone integration
Page 30
23. Immediate definition of various types of forbidden zones (like military zones, high risk industry, strategic infrastructure etc.) can be smoothly included into official USGL grid definition by authorities. 24. The concept albeit at different altitude is applicable globally. 25. The idea is compatible and extensible for more complex traffic definition using inter-drone communication in the future. 26. Geometric volume of G-class airspace below 600ft following proposed definition is in fact higher over any undulating terrain than previously below 700ft defined using whatever geometric distance sensor might measure. This provides more safety for emergency landings and more time to ground for typical G-class airspace users. 27. The idea defines safe traffic navigation rules around passages, canyons, near high mountains. 28. Because of proposed changes, all air traffic plans remain unchanged. 29. Emergency helicopters are free to operate and hover anywhere under 600ft altitude without even crossing drone-occupied band, no matter how ridiculously sparse this airspace is as demonstrated in initial calculation comparing future drone numbers with current bird population. 30. The idea is not increasing information bandwidth or Air Traffic Controller workload. 31. The idea allows reasonably easy design of drones able to fulfill requirements even in cases of total wireless communication or GPS signal outages.
Revising the Airspace Model for commercial drone integration
Page 31
4. Legal status, patent issues and future development Presented ideas have not been patented anywhere. It is important for those skilled in the art, or even those completely unskilled in several important subjects yet still able to understand basic logical flow of this document in its entirety or in any part, that presented ideas even if had innovative content, this content is now made public and its potentially innovative scope is far exceeding narrow definition of presented solutions. For example, definition of USGL and GU class airspace is technically describing an example for a method of defining traffic band for decoupling marine, undersea or aviation traffic as a function of actual vehicle kinetics and its position in space, or any other traffic or airspace property. While abundance of GPS receivers and availability of inexpensive solid state memory made this approach easier than before for flying drones, one must not overlook bigger idea behind this proposal. It is obvious and now publicly known, that any method of decoupling traffic as a function of its actual altitude, position, course, velocity (either horizontal or vertical) and other physical properties of such vehicles (including but not limited to size, density, energy, construction materials) can serve as a method of segregating space, in order to minimize general probability of collision, severity of damage, optimize costs efficiency, increase speed or just in order to facilitate independent calculation of trajectories by traffic participants to satisfy their own goals. While USGL has been proposed as based on WGS84 coordinates, in fact any grid reference, either regular, irregular, with different shape cell could have been proposed. The source of vehicle’s position is also variable: a mix of dead reckoning inertial navigation and GPS has been proposed for PSLD, but any other indoor, outdoor or mixed navigation method can be used for determining traffic participant position in the grid including optical guidance, any GNSS system including GPS, GLONASS, GALILEO, BEIDOU or similar, radars, sonars, terrain contour matching, directional radio wave transmitters, image matching, laser measurements, celestial navigation, radio navigation based on triangulation phase or other properties of the signal like cellphone triangulation and any other position determination method. Even if less efficient in certain scenarios (like islands), proposed rectangular grid seems to be logically simplest to implement. It is obvious that such grid does not have to be entirely stored onboard of generalized traffic participant: it could be downloaded or shared entirely or partially between different participants, ground stations, using any communications methods. It is possible to further optimize the idea by adjusting grid spacing and geometry to different speeds and classes of traffic participants, either by pre-computing or partial pre-computing of certain sub-parameters. For example faster and higher flying participants might want larger grid spacing, while deep sea or cave explorers conversely might want less resolution near sea surface. It is obvious that traffic participants instead of sharing their predicted path of movement, could share with other traffic participants entire grid of their current lower and upper limits of achievable space, this includes for example (but is not limited to) glide cone for gliders, parachutes, or airplanes under various classes of emergency situations. Therefore the idea is naturally extending from civilian air traffic authorities proposing unified version of USGL to traffic participants, to traffic participants sharing their own grid with other participants and generally understood traffic coordinators. The idea is obviously not limited to any specific race, planet or other celestial body, or civilian applications, but includes civilian, military and also experimental or even interstellar space reentry traffic coordination. Proposed idea extends beyond defining strictly parallel layers of altitude bands, and immediately extends into definition of truly 3D traffic model, taking into property kinetics of expected participants. The idea is immediately extendable into efficient, low cost ground proximity warning system for aircrafts, offering level of detail beyond all known aviation traffic maps.
Revising the Airspace Model for commercial drone integration
Page 32
The idea is very close to, and as it is it already forms the best available approximation to dedicated automated landing system which could have several versions optimized for different classes of vehicles. It is expected that proposed ideas might in some form invite a discussion about feasibility of global flight management systems including automated avoidance and landing. Proposed definition of grid based of its values (altitudes) at vertices is just one available mathematical formulation; besides using different grid geometry, one can represent values that are of interest to defining traffic properties in virtual space might include using mathematical coefficients or classes of functions defining the value between grid points, this may include wavelets, various polynomials including splines or rational functions or any other method of pre-computed values used to represent intermediate values. The idea of defining traffic space, as a function of positions and their kinetic properties, extends beyond representing limiting altitudes. The properties can include other factors as lower and upper airspeed limit at given point of space, noise levels, allowance for different propulsion classes, allowance of different mission classes, narrowing allowed payload class, defining airspeeds (that could be either in true required values, equivalent values or suggested values), heading, course and its derivatives or equivalent properties, allowed kinetic properties like objects kinetic energy or other definition of energy, airworthiness or more general traffic capability classes like icing conditions flight capable, visual/non visual flight conditions, environmental hazards classes (short-circuits, flammable material, soil, water or animal life toxicity) or other environmental properties that must be compatible with class of traffic participants willing to access given subspace. Certain grids can define communication requirements, bandwidth allocations, requested update rate depending on expected course and future outcomes analyzed by automated traffic participant. This way one can optimize global bandwidth of long range networks dynamically, using merged request density from entire drone swarm. It is obvious to complement proposed grid with any other grid providing additional information, like confidence intervals and regions of validity of contained data. For military applications such grid if published by traffic participant could contain mission capability factors that are of pure virtual nature, like hit probability, probability of reaching the target, maximum time over target area, payload delivery per unit of time etc. It is obvious that while proposed USGL grid is intended to be as constant as possible upon definition of every of its point by value, future families of similar grids might have highly dynamic nature depending on environmental conditions like weather, special notifications including one-time authority issued restrictions like NOTAM, temporary pollution prevention regulations and other factors that are external for generalized traffic participant. Grid geometry must not be constant, but could be recursively self-refining for certain regions. Possibilities include (but as always in this document, are not limited to) the use of well-known space compression methods like voxels, graphs and other data structures, or any compression method or any other more elaborate selfremeshing like multigrid techniques that will minimize data bandwidth, decrease volume or improve local precision. It is also possible to define traffic grid values as defined for specific part of grid, while declaring that for all other points the grid is simply obeying specific mathematical property that is preferably very easy to compute, for example initial grid proposal could include USGL model for large cities with surrounding hilltops and define that for remaining area, grid values shall be interpolated with given mathematic properties – until defined further by value. This property might dictate specific interpolation method, or a condition that global net of values minimizes certain differential equation, like 2D Laplace’s equation, leading to surprisingly immediate implementation of surface smoothing for vast, initially undefined areas, while maintaining the condition of providing a single, uniform and consistent dataset by publisher of the grid and requirement of
Revising the Airspace Model for commercial drone integration
Page 33
very fast onboard computation using limited processing power (having robustness and cost of automated traffic in mind). It is also possible to imagine very special version of aforementioned grids to representation of singular structures like obstacles for navigation, delivery locations, landing sites, recharging posts or any other not only as a grid values, but also as mostly-empty grid with a few properties being provided at certain points. Such definition is easily implementing existing list of objects, but could be also projected of any other grid. Those grids are in fact any form of mesh of any mathematical property, which might include defining entire bi-dimensional capability grid with one analytical function of special arithmetic properties, effectively reducing grid content to a few coefficients. One can image more than a single copy of given grid to be stored on different media types either onboard or on ground-based system controller locations. This kind of optimization can represent either the same grid stored for different access time, with different area coverage, or different resolutions or geometries of the same grid, either for redundancy or computational efficiency purposes. It is worth mentioning proposed grid approach is compatible with parallel computing paradigm, when several computational entities (multi-core processors, collaborating groups of automated traffic participants or redundant computers onboard any of mentioned vehicles) are evaluating possible outcomes based on shared data. One can also immediately imagine applying logic values at nodes of mentioned grids, like no-fly zone, special air traffic areas or other as described by the law. Defining specific grid type and interpolation method, one can represent geometry of circular no-fly area around any given point by binary values over mentioned generalized grids that would also define computationally simple methods of determining the idea of proximity to certain areas. Logic values, once interpolated on a grid of carefully crafted geometry, can yield as a result fractional values, creating ‘proximity alert’ functionality, that could be included as weight factor to general flight planning onboard automated traffic participant willing to either to avoid or to prefer traveling in certain areas. Such smoothed representation of automated traffic areas are compatible with neural network-based navigation algorithms, quantum computing or stochastic navigation model which often increase decisional flexibility if other-than-binary representation system for logic values is being used. It is obvious for general public that any grid defining partial transit or travel capabilities does not necessary have to be made of values representing the same range or resolution. Application of so-called dynamic range is obvious, corresponding to variable resolution of information represented by such grid. Furthermore, values represented, when applied to airborne traffic, might be significantly simplified if a traffic participant crates his own copy, where certain positions like a list of predicted destinations, takeoff locations, registered or amateur airports emergency landing sites or most valued customers are pivotal points where grid value is represented with maximum resolution, while all remaining points contain information of various resolution, degrading with distance from those sites, or are defined with a compression method that favors fastest computation at those pivotal points. Since several drone flights might be expected to return to their takeoff site, one can easily represent reference elevation grid as set of points containing relative differences to its neighbors, which for most surfaces will be close to zero, therefore easily compressible and represented by lower number of bits. This approach, unlike general data compression algorithms which are based on stream data without caring about their internal meaning, could lead to significantly reducing onboard memory and calculation cost and therefore drone price by using the information about geometry of onboard grid before flight. Therefore, one can profitably compress mentioned grid data using its context, which is geographical location of selected important places.
Revising the Airspace Model for commercial drone integration
Page 34
It is obvious that any mesh or grid used for defining traffic band or sectors in space that among other things could serve as reference definition of airspace, obstacles and permitted behaviors, can use variable data density based on irregular grid spacing, different definition of interpolating function, interchangeable use of data representation by value or by coefficients of any mathematical function or by other mathematical properties like cross-dependency of values between nodes, directional or content-dependent compression algorithms, different geodesic reference (datum, planet’s ellipsoid), different reference system (like the one presented, semi-Cartesian based on latitudes or longitudes, any projection method, radial based on distance and heading from specific point like takeoff place or any other) or dynamic grid resolution or use of different (triangular, rectangular, hexagonal or any other) meshing elements, in order to reduce computational costs, power usage, automated landing precision, better obstacle avoidance in certain areas or any other goal devised by its users. It is obvious that among other data content represented by a mesh or grid used for traffic regulation and assistance (automated flight planning and execution in particular) might include environmental data or data assisting in finding emergency landing sites in real time, also for manned aviation. Such information can include data about natural conditions like rocky or undulating terrain, presence of trees, swamps, several unlisted man-made structures like walled rice fields, population density or time of expiry for information contained for representing areas with expected changes, particularly after natural disasters, with fast growing trees, regularly flooded. This data can serve as emergency landing air traffic aid, while its validity can be regularly tested by several drones which being much shorter-ranged, frequently landing and operated by geographically more localized specialists, will be serving as testing community for validation of environmental emergency landing data. It is also obvious to extend the idea of sharing transport capability grid between recipient, general traffic participant(s) and traffic coordinator(s). For example, a delivery drone knows its capabilities and measures meteo conditions in real time, publishing its grid-based values. The grid itself might be centered on traffic participant (in this narrowed example a drone) and capability might be arrival time to given point of space. From client’s perspective, one can have additional benefit of this approach: one can ask for delivery times of goods using a drone around his location indirectly, without revealing his own location or while anticipating movement prior to delivery. In such mode of operation, time to delivery to any specific point is not published, protecting recipient’s final destination and privacy. Furthermore the grid requested does not have to be related to precise actual position of a drone or recipient of information. What is published by drone is average collective system’s time to delivery for every point in the area described by mentioned capability grid.
It is obvious that such capability grid describing traffic
movement parameters or restrictions does not have to be mobile or attached to any single user. A drone carrying payload can arrive to pre-designated point, just to receive general broadcast of new set of users demanding delivery. This situation update could be repeated several times and the drone is operating on fusioned data, without ever being attached to any single destination either final or a milestone. This technology would allow ordering a drone delivery while traveling in a car, while looking for empty field or parking lot for final position update just before final goods delivery. Obvious generalization of time of delivery is time to arrival to any other intermediary stage, like hovering or loitering drone that is in ready state for immediate delivery at requested position below. Instead of displaying at client’s devices (which may be dedicated, phone, general computer or any other delivery management device), one can easily imagine a trivial system that extracts time to arrival to any of intermediary delivery stages, subsequently waiting for the recipient to refuse delivery or to proceed to next stage of delivery, which might include airdrop, landing or another kind of deployment. Because delivery is optional and the time shown is only to intermediary point, displayed time to
Revising the Airspace Model for commercial drone integration
Page 35
intermediary navigation point prior to final decision is not equivalent, but statistically mathematically close to time required to complete delivery procedure. It must be noted that this staged action is not restricted to narrow idea of delivery, yet whenever we think of delivery we can also consider picking objects, counting them, exchanging object one for another, replenishing stockpile, simple building works, piling up objects in organized manner, replacing mistakenly shipped objects, replacing damaged objects, finalizing previously incomplete object deliveries because of vendor’s fault (a.k.a. FlashReplace service), offering reduced-time complementary deliveries (additional yet essential parts) for certain complex products (FlashSupport service) but also area works like randomized disposal of chemicals or pellets (f.ex. pest and predator control), spraying, scanning information, surveillance operations, refueling, retrieving data using short range high-bandwidth links, making photos or other recording media like sound, video or other scientific measurements with geological application among other things, or performing any other mission offered by drone operator like Commercial Search And Rescue (CSAR) for malfunctioned drones. It is also obvious that information contained in such traffic capability grid or mesh is not restricted to any single variable, as it could contain hybrid data like payload delivered per unit of time or mission time over specific location. In more general sense, proposed grid formulation in itself is an abstract layer for navigation, including delivery drone navigation, that by design omits restricted or otherwise unexploited space. It is therefore not proper to say that any device navigating within his grid of capability will be following GPS or any other coordinates, as vehicle’s awareness about its position will be entirely related to grid system. Certain specialized position description systems do exist, MGRS being an example, but for purposes of delivery and logistics other geometries including radial or distorted binary radial coordinate systems an be imagined. While USGL is using angles defined over virtual ellipsoid for defining node positions, same as used by GPS system, one can imagine using any other grid system following to some degree one of known projection methods used in geodesy like Mercator projection. In this projection all lines of constant bearing are represented by straight segments on a Mercator map. One can extend this idea by creating custom projection, evolving with time and published within drone navigation system. For example, said grid could be triangular mesh, with triangle sizes varying with dangers of flying over given terrain. In other words local mesh density could represent any navigation factor to be taken into account. Using such grid, by drawing straight line between any 2 points, an ensemble of mesh elements (or triangles) could be immediately selected using simple algorithms and Cartesian geometry. A chain of said elements is forming approximately straight line in grid coordinates, but would remap into optimal, curved flight path in real world coordinates, avoiding as much danger as possible. This way one could evaluate almost instantly any number of paths that are automatically taking into account most cost/risk factors. One o possible applications would be eliminating the effect of artificially created high-density traffic areas or corridors that are automatically appearing at the edges of airspace zones that must not be crossed or large obstacles or clusters of said obstacles. An example is artificially high concentration of aircrafts at the borders of present day CTR/ATZ/MATZ areas, as some aircrafts want to stay within area directly supervised by given air traffic controller, as well as artificially high traffic right outside along the borders of said CTR/ATZ/MATZ zones because of general traffic aircrafts trying to avoid forbidden area at minimal cost and effort. In case of generalized projection grids indicating general area to fly by right at the moment of plotting virtual straight course in said grid coordinates, would immediately uncover grid cells that are larger where flight path can be relaxed and where it must be kept tighter in statistical, non deterministic sense, allowing very easy implementation of smoothly spread traffic.
Revising the Airspace Model for commercial drone integration
Page 36
It would be naïve to restrict proposed averaged grid paradigm to drone navigation to be restricted to two dimensions. In fact, proposed USGL grid is already a projection of 3D data on 2D grid, which in itself is projected via angles(latitudes and longitudes) onto earth surface. Therefore mentioned grid approach is not only extending into spatial, temporal projections (velocity vs position data) but also might in fact correspond algebraically to fractional dimension of data (where certain movements are degenerated to two or less dimensions, while other movements in certain parts might allow more degrees of freedom). The invention is hidden in ability to renormalize navigable outcomes into geometric systems where data shared between traffic participants and supervising centers are merged in uniform way. It is also improper to say that any device finding his position relative to grid of capabilities will be aware of its destination position in deterministic sense. The most space-energy efficient description of traffic involves variable dynamic range of information per cell divided between object position and velocity. This concept is inspired by quantum mechanics uncertainty principle, which applies very elegantly to traffic participants: those who move faster should have less defined position, therefore occupying statistically larger space volume for purposes of navigation avoidance, but also for purposes of prediction of their movement. Similarly, a client ordering a product should not emit his position, rather a smooth combination (using any mathematical formula) of his position and velocity vector that depends on his own measured velocity. Such approach offers smooth and coherent transition from delivering to fixed position to transit delivery following general moving or stationary delivery target. There will be no state change as seen by delivery drone or other traffic participant trying to minimize his own position described in such position-velocity space relative to destination, which loses his geometrical fixed meaning. Further flexibility is related to defining delivery request as blended representation of direction and distance from source. Far away, more bits of information will represent direction, while at smaller distances more precision will be devoted to distance. This information being updated in real-time, kept on a density grid of requests for given type of delivery or more general mission request, would protect privacy and location of requesters and allow automated dispatching based on request density rather than treating any request as a single event. In this sense any flowchart defining time to arrival, in essence based on stationary geometry, is being superseded and generalized by proposed system merging both velocity and positions. In such a system there is no time to arrival, but rather time to merge position-velocity mix, which (while being quite close numerically in many cases) is degenerating to a very special case of matching delivery position only in case of perfectly stationary delivery location that remains motionless from point of view of navigation system for infinite time, with infinite precision of position measurement (no drift). This formulation based on capability grids of traffic participants represents a non-deterministic system that is not based on sharing exact position of anybody, until they become perfectly motionless for infinite period of time. Such proposed adaptive traffic management and navigation grid system is therefore extending definitions of scheduling, data sharing or implicitly using Newtonian motion dynamics as in previous proposals in navigation. Proposed generalized idea of delivery time, range or other capability expressed in mentioned grid of probability transmitted between drones also includes guided or unguided multi-staged payloads, each with different autonomy class and different level of safety for traffic environment, thus having greater abilities to penetrate different space subclasses meeting required safety levels. The idea of navigation is also completely replacing takeoff-delivery-landing paradigm, with drone logic being endless stream of sub-decisions, some of them might be limiting done capability matrix as seen by drone itself, being aware of remaining payloads, propulsion and energy. This generalization allows both flight recovery (if single-use recovery mechanism is triggered but then situation fully recovers), as well is not cutting data or
Revising the Airspace Model for commercial drone integration
Page 37
storing exceptional events away from general stream of decisions. This property alone will make system highly flexible and applicable to rapidly changing drone technology. In that sense it would be easy to make a drone that completely shuts off upon arrival or after detecting any major system malfunction, just to be smoothly restored afterwards without entering any obscure or rarely tested code. Obviously, the idea of single target flight, or workflow consisting of target selection, routing, takeoff, action, return routing, landing is completely obsolete, as practical application of presented ideas might result in statistical sense in abandoning then resuming mission several times, multiple delivery per flight as understood in classical way of thinking about aviation, and multiple flight or even takeoff and partial recharging attempts seen as continuous stream of drone decisions, optionally broadcasted to the system for improved drone collaboration based on averaged demand/environmental grids, Naturally, sharing generalized time-to-target or similar capability grid data between automated traffic participants can allow advanced traffic optimization and rerouting techniques, like scheduling a different drone by itself to a new target increasing delivery probability, swapping payloads, or swapping targets for two or more drones in flight in such a way that given changing weather conditions taken into account by drones themselves, the procedure will minimize delivery, time reserves, transmission bandwidth usage, select different subset of emergency landing sites or regular landing sites, optimize consumable resource usage or other factors like environmental impact. The difference between classic route planning is that re-routing will not occur based on source-destination route planning, but rather on merging sub-sections or pre-defined subactions in non-deterministic way, that merely prefers drones to deliver, but much more invites them to return home at the edges of their navigation range/capability grid. It is obvious to the public that sharing capability grids instead of requesting drones to enumerate capabilities to reach each target, recipient or point of space by automated traffic participants is decreasing computational workload of those participants, as they accept constant computational effort publishing those grids, rather than enumerating unpredictable number of congestion-prone evaluation requests. Purposeful side-effect of mentioned technology is natural transition between areas served between different logistic centers, as there is no centralized drone management, only broadcasting of drone decision at irregular, statistical intervals that is evenly using scarce transmission bandwidth. Total or partial merging of broadcasted drone decision data could be performed by any ground site, progressively increasing drone fleet awareness as information from remote areas arrive, being retransmitted to local grid merging posts. It is natural in the context of navigation of traffic participants or any other delivery drones to evaluate onboard any number of possible short-range outcomes, each by itself being a complete and safe navigation sub-path along with its initial and final state (delivered, battery depleted, crashed), that will be evaluated in parallel and selected based on cost-function which is a result of evaluation of several others cost-grids, that include mentioned previously grids indicating violation of airspace of progressively large consequences, or even larger numerical costs of crashing over populated area due to depleted battery, or smaller costs of not meeting delivery time of a system of drones as a whole. In that sense, formulation of movement changes from Newtonian path prediction, to another general formulation of dynamics similar to extended Hamiltonian, which may include selection of drone path or action among several solutions and choosing the one minimizing energy of a system. Such approach is vastly superior to participant-focused management, as it automatically adapts to any level of congestion without adding any transmission, calculation or transmission workload per participant to the system. Another obvious idea is to use genetic algorithms, having a few breeds of possible combined path+action sections evaluated by onboard computers, recalculating their energy contribution to the global system,
Revising the Airspace Model for commercial drone integration
Page 38
choosing the one set of path+action minimizing energy to be executed and proposing a small family of future paths to be evaluated that are semi-random or deterministically improved mutations of previous bestmatching sets of path+action. If we add elements of statistical dynamics to the path in form of temperature, that might be implemented by ‘hotness’ (meaning high probability of changing state, in this case a subset of path+action) of traffic participant due to its capability to navigate slow or only to very close distance, we can create navigation system where a layered grid system serving to evaluate several factors impacting energy of delivery system as a whole can be both partially updated – in presence of telemetry, or can rely solely on changing perception of traffic participant, like changing meteo conditions thus increasing cost to arrival along certain path. Given high virtual temperature of certain participants (for example those who know they cannot continue flying much farther in the storm and they are still over small yet populated area, and another last known transit capability grid has large numerical contribution to global energy because of merged velocity+position quantities just on the path ahead), some drones or any other participants might chose with higher probability to drop their willingness to satisfy their payload density over remote delivery cell. In presence of feedback telemetry, this information will be absorbed by the system as general inability for entire system of drones to increase payload density for delivery cells somewhere on capability grid describing payload delivery and progressively broadcasted to listening traffic participants. This scenario can be seen by many of the drones waiting to takeoff with desired payload while analyzing constantly their next action. At some point, taking off with partially charged battery but correct payload is by luck tending to significantly minimize global Cost Function, generalized Hamiltonian or any other mathematical formula just by defining next path+action for one of the drones to detach and takeoff. This system has very significant advantage over existing flowchart-based rigidlogic formulation. While traditional deterministic decision process described by flow structure looks neat, its implementation taking into account any number of participants in any complex yet deterministic algorithm, creates heterogeneous, impossible to maintain code, which must switch between different stages of severity of navigation which can contain hidden bugs leading to abnormal and seemingly random behavior. The result would be complex, hard to maintain logic based on deterministic formula that would not provide safe and deterministic looking system for external observers or ordinary operators. Proposed solution naturally makes any patents or patent applications based on deterministic flowchart immediately obsolete: a solution proposed is only a few times more expensive computationally than classical single-path planning, yet supports infinite number of factors like prices, availability, resource, transit space sharing and other factors in a neat few capability grids, collision avoidance, among them USGL as safe ground plane for determining altitude band, for the final outcome of uniform drone swarm code management that can be trusted. Corresponding to this proposal logical decision diagram, if ever drawn on a single picture, depending on granularity that is prediction length and scope of each path+action, would rather look like fuzzy image of infinite number of flowcharts some going in opposite direction (return to base) others changing state in any direction, making any existing process description of existing drone navigation system as merely very special case, that is equal to the final outcome of presented navigation system, but only with very high probability. Once properly set up, proposed non-deterministic system is naturally accepting introduction of any new traffic or action element in the global pool of possible actions, as there are no hard bonds inside navigation code between landing and takeoff or any other action. Indeed, as the wind for one drone changes, one drone might be nearing costly and uncomfortable decision to continue or ‘have some backup’, but it is not requesting it directly, rather by publishing its mission capability grid. Its low chance to finalize mission success make this drone ‘hot’, willing to change course, while other ‘cold’ drone will be very unhappy to takeoff and support the other drone
Revising the Airspace Model for commercial drone integration
Page 39
because of having desired payload, but in quantity larger than ordered. As time progresses, it might turn out that the system based on individual drone’s decisions converges to backup drone taking off to support the other one, juts to realize that the wind has changed again and the supporter has similar chances of delivering payload, but with incorrect payload quantity. Followed by carefully selected temperature vs decision change probability could lead the supporter to decide to land again, while the original drone to continue to ‘risk’ delivery having only 90% of chance, seeing now slightly improved meteo conditions again. Such smooth situational management while being reasonably close to expected and optimal behavior, using deterministic code could lead to insane complexity of deterministic logic (ability to interrupt charging, decision about delivering of surplus vs non-delivering anything), Consequently described functionality would most probably never been achieved because of high cost of testing real-life flying drone systems. In other words proposed navigation paradigm not only supersedes deterministic navigation technology by itself, but also makes solutions from patents below an outdated representation of decisional logic, where presented scenarios are just a fraction of possible outcomes and navigation methods to which an ideal system would converge in perfect conditions: US2-05/0277440 A1 Sense and avoid for automated mobile devices (proposing an algorithm that involves real-time analysis and sharing of (N+M)^2 possible interactions, where N, M are number of obstacles and traffic participants, with no backup strategy for fail-safe scenarios, no provision for handling of environmental impact or usage-restricted zones) US 2016/0039529 A1 Propeller safety for automated aerial vehicles (proposing unidirectional detection action outcome, instead of probability of detected occurrence total cost of the final outcome) US 2015/0120094 A1 Unmanned aerial vehicle delivery system (revealing directly positions of all actors and rigid logic and risk management structure, no distributed decision processing including global system awareness is possible, hard to maintain conditional dependencies in mission-critical code, very bad scaling properties, very bad impact of local congestion on processing, critically dependent on data link, assuming constant precision of position in space for planning, unnecessarily complicating logic by differentiating between stationary and moving target delivery) US2016/0239804 A1Modular air delivery (revealing directly positions of all actors and rigid logic and risk management structure, with centralized decision making vulnerable to taking any reasonable decision in case of transmission outage, rigid point-point transport mission planning prior to takeoff, note: this patent application is in large part invalidated by presentation [PPOST] because of prior art dated 2013-02-13) US 2016/0196755 A1 Ground effect based surface sensing in automated aerial vehicles (assuming static logic for determining applicability of entire procedure, without integrating any risk management of non-functioning of mentioned method into general navigation plan, functioning requires knowledge of presence of all other traffic participants above landing pad that is numerically costly to evaluate and in practice unreliable) It is obvious to general public that mentioned in those patents/applications missions and tasks can be rewritten and solved more efficiently using more general capability grid, energy-based non-Newtonian navigation planning using stochastic processes which include proper resource, safety, privacy and redundancy management at decentralized level. One must understand that mentioned documents are public documents, and in areas when they fail to propose a complete solution, one can propose public, complete, coherent solution that is explicitly crafted in such a way, that it is designed to successfully cover a system that is
Revising the Airspace Model for commercial drone integration
Page 40
equivalent or better in performing all actions and applications, either explicit or implicit, mentioned in patents/patent applications mentioned in this document or any other document related to automated traffic that are available to the public at the time of writing. Further generalization of delivery property is a blended mix or multivector of position, velocity represented by demand grid similar to proposed USGL grid. Such grid-based system using non-deterministic logic and dynamic range of precision will naturally blend delivery systems or any other service system to dynamically incorporate various requested time to delivery. In other words, proposed system of demand representation will allow weighted integration of different prices of products, prices of delivery, position of delivery along its expected physical velocity forming a system that prioritizes all available resources guaranteeing timely delivery of maximum number of items or supporting services. While all those calculations are not leading to deterministically optimal solution in all cases, they are avoiding analyzing each case in deterministic sense what would lead to extremely costly NP-hard class algorithm making entire calculation pointless beyond the simples cases comprising of a few dozen drones. In practical application, it is possible but suboptimal to offer say 30 min delivery and deliver in rare cases in just 5 minutes, the reason being that specific time of delivery is requested by individuals or organizations in order to meet their other plans or deadlines. Such occasional ultra-rapid delivery of services or goods would raise expectations, disrupt other scheduled actions which might be already under heavy decisional workload and lacking the time to adjust their plans, could create false impression of repeatability. Occasional rapid delivery would make more difficult to assess risk management options for delivery/logistics center in case of partial local shutdowns or delay due to unrelated service, as rushed delivery drone is unable to serve as backup for other missions. Besides, such generalized demand representation allows natural inclusion of various classes of transport system participants, delivering at different guaranteed delay like 30, 60 or 120 minutes. From this perspective, time-to-delivery represented in some patents is simplistic, primitive subclass of desirable outcomes, which consumes ill-scaling amount of resources for drone decision making and is incompatible with optimal functioning of delivery network as a system. It is now obvious to general public that existence of proposed dynamically prioritizing delivery system would allow business scheme that is properly taking human desire for urgency in proper way, including different prices of delivery that do not necessarily need to be fixed. It is easy to imagine a system that manages services or deliveries in such a way, that online or other money or equivalent value based system is allowing bidding for accelerated delivery, presenting system response time for specific bidder delivery demand. In extreme case a high bid could cause rerouting several items already airborne to most demanding customer, at minimal time possible. On the other hand nighttime rural deliveries that could be scheduled in advance could lead to serious delivery price rebates, properly equalizing available drone fleet usage. Obviously, proposed invention while being surprisingly easier to implement than to understand by its capabilities, is rendering ideas patented above completely obsolete because they fail to implement key elements of demand-driven economy. One of the possibilities would be to offer smoothly evaluated bidding system causing drone to increase its willingness to pursuit to its delivery zone facing unfavorable weather for high-bid deliveries. An old idea of displaying time to delivery based on flight parameters of any of traffic participants is essentially making the client psychologically dependent of several additional factors like weather, which should not be the case. As with other traffic or mass event systems, timely delivery meaning no premature deliveries complemented with sincere real-time estimates of delays of entire delivery system is a key to establishing trust between clients and delivery system operator.
Revising the Airspace Model for commercial drone integration
Page 41
It is desirable for a system to be able to naturally merge into general order systems with clients desiring to maintain maximum secrecy for personal reasons, just to mention a possibility of secretly ordering Rap music CD while living in some highly eccentric and conservative society of philharmonic orchestra musicians. One would want to be able to place an order that could not be decoded into his own personal whereabouts by entities less equipped than security officials tracking bank card numbers or other data that is otherwise very well protected even in case of massive leak to the public. It is interesting to consider proposed position+velocity combo including other delivery or transportation factors calculated based on several grids as property vector that can serve as generalized senor and awareness fusion. It is possible for example to represent measurements that could not be represented otherwise while satisfying sensor properties, for example Doppler radars or any other devices that can measure velocities of moving objects in certain direction with greater accuracy. It is also possible to consider a possibility to order goods by a user that is unwilling to share his current position because of intentionally disabled positioning system, but agrees to share his own speed or heading, bus number (those data could be embedded in a car or any other current transport vehicle or entered manually) and a hint of his temporary position indicator like county code, cellphone tower ID or internet address (MAC, IP or other) allowing partial position estimation based on wireless networks. This would ensure faster response of a drone to demand by flying to approximate area with more precision position estimates to be entered as a general probability of delivery at better defined position at later time. Further development of supply-demand grid used for navigation of automated traffic participants for increased privacy include sharing with ordering party one-time geometry grid, which can be stored on one or a few delivery devices carrying desirable goods, without being shared everywhere else. There are also other concepts of one-time grids, that can be have various levels of volatility/permanence, depending on regulations and national safety concerns, for example restricting type of commodity being delivered with such a method. Such grid can contain terrain or other geometric features known only by the offering and ordering party and a list of features can be generated by the offering party progressively until the customer identifies delivery position or a mix of velocity+position on mentioned grid. A geometry of each of those grids can be encrypted with any useful algorithm, making participants sure that even total teardown and disassembling of flying device will not reveal neither initial drone position nor expected delivery position. This approach doesn’t prevents other automated traffic participants to fulfill the same mission as long as they share public encryption key or other method to decrypt the grid for their own purposes. It should be kept in mind that any grid geometry can be used, as mentioned everywhere in this document, what also can degenerate to 1 or 2-point delivery grid describing a single or any pair of locations or quasi-location being blended probability of position and velocity and other good delivery parameters. In retrospection, a definition of decentralized personal computers that were not necessarily connected at all time was key element that allowed for explosion of applications and technologies in late 80’ in the past century. In case of computing, outdated centralized computing approach was surpassed by personal computers because of greater efficiency, scalability and overall higher performance per unit within a model where several independent entities could coexist following data sharing standards. All this resulted in relatively minor dangers, none of which reached level of rebellion nor eliminated the liberty of calculating things manually. It would be unwise to design any drone management system based on centralized processing while proposing multipoint-multipoint network like Internet as it would reveal total misunderstanding of entire Internet concept. Said network was designed specifically for avoiding having any single point of failure in heterogeneous network and instant fully transparent routing between network nodes, immune to failure of any number of nodes in case of limited nuclear strike. It has gained popularity because at the same time the
Revising the Airspace Model for commercial drone integration
Page 42
system was immune to human factors, local failures, misconfiguration, natural events and many more. Using heterogeneous networking paradigm like IP protocols in order to distribute data managed by strictly centralized processing power would be total misunderstanding of data resiliency, mission-critical computing, almost all modern computing practices and the history of Internet. Besides all technologies mentioned, flying vehicles performing delivery of goods, or any other automated traffic participants, will not loose their identity in supply or service management chain. It is obvious generalization of packaging concept to label drones as containers, their subdrones or payloads or payloads of said subdrones (ad infinum) as logistic containers or generalized packaging, that deserve proper labeling with any practical method, including shipment ID, barcodes, RFID technologies and any other method that is among other things applicable to automated identification, scanning and inventory listing systems. This information does not have to be limited to invariable ID of a drone as a fixed container entity, but also a configuration when multiple container IDs of which most can be empty will be permanently affixed to transport participant, and only the one that is virtually filled depicts specific version of drone configuration that is actually flying. This way we can include configurable automated transport participants into existing supply chain worldwide. Marking of a drone or generalized automated traffic participant does not have to be affixed to transport aircraft. It is possible to apply existing automated sticking/labeling technologies to airframe body or any other part separately for each flight. Keeping in mind that disassembling things is usually easier than assembling and that flying objects must be checked in their entirety before their flight, it is reasonable to assume operation mode for delivery drones when instead of being assembled from interchangeable parts on demand, a generic, Fully Functional (FUFU) flying drone is supplied by default to logistics site, then it is partially dismantled disposing off unnecessary elements for expected set of mission profiles that given day (we are not in single drone-customer-mission paradigm anymore!). Any missing components removed can be supplemented with on-site manufactured made on demand elements in case of detecting their failure or unsuitability, making seemingly fixed drone design able to be customized for a mission in spirit of general safety while eliminating any assembly process on-site, with obviously positive impact on flight preparation time and reliability of what is left for flying. After dismantling elements like excess additional power supplies, sensors that are useless for given mission profile, excess transport pods, propulsion units or unnecessary aerodynamic surfaces offering superior reliability of actions that were intended to be covered by mentioned patens about modular air delivery system. Mentioned process is essentially superior and reverses assembling Modular Air Delivery (MAD) which are prone to assembly mistakes and is smoothly replacing the concept with more efficient Partially Dismantled Vehicle Operations (PDVO). It is non-obvious compared to other technologies, like PC computer, when one may ask for a specific configuration from ground-up and wait for the result. In mass manufacturing, the opposite process of intentionally slightly over-designed system is preferable, as this ensures that assembly process which is more demanding is applied at lower labor/energy/price/specialization effort than assembling a system on demand anywhere on the world. Obviously unused components are reintroduced into automated traffic participant manufacturing chain so that nothing gets wasted on the long term. During preparing such drone, one should not need to take into account expected single mission profile, as in fully cooperative navigation system this is undesirable, but rather by making system suitable to generalized subclass of delivery missions. This family of average mission profiles requested can be extracted from proposed demand/range/price of delivery grid system. From this perspective, proposed technology supersedes patent US 2016/0239804 in all its capabilities, creating specialized fleet of average properties on demand out of proven, fully assembled generic
Revising the Airspace Model for commercial drone integration
Page 43
drones by subtractive process, making fleet more reusable and fit for in-flight cooperative role interchange as described by proposed statistical energy-based swarm management paradigm. Further expanding the concept, drone specialization for selected group of mission profiles might include variable geometry of components, be it fuselage sliding relative to wings, payload pylons/containers moving inside or under fuselage or wings, or variable wing and propulsion mounts geometry (twisting, turning parts, sliding in and out, scissor-like folding, tearing, inflating, deflating) corrugated rails for sliding payload (similar to rails used for mounting aiming aids to firearms), or adjustable electric engine geometry changing its power requirements and efficiency profile, or adjustable propeller geometry like angle of attack of its blades or any other geometric change that is not requiring dismounting anything. In this concept, vehicle capabilities form a continuum and their capabilities are smoothly adjustable. This will naturally create as many types of vehicles as total number of traffic participants making centralized management unrealistic by other means than mentioned capability grids, that are uniform concept of sharing vehicle’s own mission capability evaluations, which can change during flight as a function of detected or downloaded conditions, also including history of device’s component like battery or general efficiency of propulsion. In narrowed and strictly applied version of PDVO philosophy, key components remain associated with given drone until replacement at their end-of-life, what means in practice more accurate history of usage tracking allowing better prediction of remaining flight time or propulsion system status. Another obvious idea is to fit battery supply systems with their own memory of usage, current draw, capacity and even entire discharge cycles for better component supervision, however only automated vehicle itself can include all factors including statistical usage history of its neighbors that is worth sharing with other decisional entities within navigation system. Wearing factors might include additional cooling/heating due to adjusted vehicle geometry that can affect expected performance of a battery for given flight at given meteo conditions what might be know only to dedicated, fixed flight controller of a given drone. Once again publishing partial technical parameters across entire system appears to be inappropriate for fleet management, while smoothed grid sharing system comes to the rescue. Proposed variable geometry is not requiring assembly of components, but rather adjustment of their relative placement what is affecting very important aerodynamic parameter Center of Gravity (COG), Center of Lift and Center of Thrust in all flight regimes. In practical application one might imagine pre-flight COG measurement which requires placing a drone on designated measurement device called Gravity Measurement Post (GMP) comprising of at least two, but possibly three or more pressure sensors. Knowing geometry of given device, GMP can use two or more pressure sensors (either two placed under two bars/plates or any number equal to ground contact points of given device) in order to transmit to vehicle its actual COG position. A vehicle can respond to system with its own desired COG or propose adjustments to be made (like pivoting wings, engine mounts, payload, power source, removing or adding certain elements). Such changes might be evaluated by ground system or manual operator/technician who having only one of a few classes of overdesigned FUFU vehicles can propose necessary geometry adjustments. Whatever changes were made, actual COG position can be then transmitted to vehicles onboard systems in order to facilitate automated takeoff and optimal performance of flight control loops since the fist milliseconds of flight. Transmitted data can of course include All-Up Weight (AUW), which is related to inertia forces impairing steering agility; therefore this data is vital for improving dynamic takeoff capability or landing precision. Furthermore, it is obvious that mentioned Gravity Measurement Post device could assist on verifying pre-flight weight distribution even in case of specialized, non-modular delivery drones. One can easily imagine drones specialized in carrying one shape and weight range of an object, be it A2/A3/A4/A5, Letter, Legal or any other
Revising the Airspace Model for commercial drone integration
Page 44
paper/book document format, or CDROM, brewery and food containers in any of a few standardized sizes, or entire cellphone/PDA line of several manufacturers which could be carried either in specialized cargo bays or inside relatively flat aerodynamic surfaces. The idea extends to any predetermined multiplicity of said objects and can also include objects delivered by their manufacturers according to logistic center specifications that allows optional air delivery, so-called Air-Ready or Drone-Ready packaging. Such packaging could be made of any material normally used for packaging, but would include foldable hinges, hooks, latches, surfaces, prefabricated edges and optionally tearable side that would reveal special geometry facilitating drone delivery. One can also imagine strictly specified drones carrying data storage devices like Kindle or AirKindle or DroneKindle book readers, music players of various manufacturers with names starting with A, M, S or any other letter. When it comes to letters, drone shapes might be instrumental in enhancing brand perception in general. One can imagine delivery drones shaped as certain logo, digits, symbols or letters, be it Latin, (letters A, W, X, D, B, H, S, U, V, Y, M and T show the most flyable potential, L being limited mostly to certain tribes in remote lands) Arabic (for evasive maneuver classification), Cyrillic (for military deliveries), Chinese (bulk deliveries and meditation equipment) or Greek Alphabet (letter Delta has important flight performance and historical significance). In certain countries one might want to avoid certain runic symbols, even if they would present a few interesting emergency recovery aerodynamic properties. This idea would inevitably associate any mishap with given brand, but this is inevitable in case of any accident investigation. On the other hand, this idea is enhancing public perception and feeling about drones, also a sign of responsibility, as depending on operator and judging drone shape one can naturally associate it with positive (goods delivery managed by professional operators) or negative reactions (foreign military intervention). Whenever outside the container thinking is applied, because a drone with payload is heavier than without it, therefore one can imagine packaging prefabricated in such a way that is reinforcing part of drone structure it is attached to, in order to compensate for additional load of said structures when carrying heavier overall weight. This is drone-aviation counterpart of integrated fuel tanks and could be called Integrated Payload or Integrated Container concept – in the sense there are no container in itself as much as there are no fuel tank in integrated fuel tank concept – just airframe structure that happens to contain something special. Those devices similar in shape yet much more advanced in nature and role than classic containers could be fixed in shape and totally non-modular, differently attached to a structure (for example at different points along the fuselage or aerodynamic surfaces) depending on expected flight profile or weather condition, but for duration of flight they would effectively form an integral part of a drone. From this point of view detaching such structurereinforcing container in similar ways as in Modular Air Delivery patent application, is in fact not just detaching a container, but rather extending a concept of partially Disassembled Vehicle Operations (PDVO) into Self-Disassembling Vehicle Operations (SDVO): a vehicle can decompose itself in a safe and controlled way increasing its mission capability. From this point of view, not only a pre-charged power source, but also a power source that is in fact part of the airframe can be picked or dropped somewhere along flight path in order to increase remaining endurance for remaining drone setup. One can also imagine Designated Emergency Drop Area, (DEPA) used in case of non-delivered payload that is still part of a drone, when a return of a drone is judged to be more important than risking to return with heavy structural drone element attached. That way, one can imagine extending vehicle range simply because of case of non-delivery, there will be guaranteed Designated Emergency Drop Areas along the route, which guarantees that one can extend drone range delivery capabilities as if half of the flight would be without some significant weight, regardless of the outcome of delivery. One can imagine multi-staged vehicle decomposition, leaving in controlled way takeoff
Revising the Airspace Model for commercial drone integration
Page 45
assist parts or undercarriage, part of propulsion systems or power source on the way out, then structural part containing payload, then loosing unnecessary landing assist sensors, or any other sensors, including unnecessary flight emergency devices, then another part of power source along flight route in various combination depending on normal operation or progressively higher severity class of anomaly. This way, instead of charging any device, power packs that are either separate or being designed as structural part of drone structure could be left for charging at remote sites, at which power pods might be picked or left for charging. It should be noted that while within proposed framework a concept of takeoff or landing is neither necessary nor unique event in daily chain of events, in practice one can observe as an extreme application of SDVO disposing off unnecessary parts from a drone itself before takeoff as a result of last minute environmental or traffic estimations, or disposing off certain structural elements on remote locations which might not be unique part made for one class of disposing drone, what can lead to creation of Replenishment Outposts (RO), Payload Exchange Docks (PED), Excess Payload Drop Area (EPDA) or Designated Emergency Drop Area (DEPA). Entire system can lead to creation of multi-staged Relay Delivery System (RDS) that could be made extendable geographically in certain days, directions or special conditions like emergency situations support. It should be noted that said drones that are being assembled, adjusted or disassembled might contain production parts that are elements of their structure while on the ground, but are removed at the end of disassembly process just before takeoff. This could create a situation that factory-proven drone is devoid of certain components not only useless for certain mission profiles, but also those that are only present for practical on-site handling purposes, for example reinforcing the airframe during flight preparation process. Further extending the concept of SDVO, one can imagine Transforming Vehicle Operations (TVO), in which a drone or general automated traffic participant is a product itself. It could be that airborne delivery of drones can be made exactly that way, but the idea extends where any part or majority of flying drone is made of ecological, decomposing, discardable or returnable via classic mail elements. Final user/client himself can dispose off unnecessary parts transforming received product into desired object, potentially returning to logistic center any unused parts at later time. TVO drones can use product’s own battery or other parts, but also TVO body can be made as intentionally disposable, consumable (by humans, pests) or returnable to Drone Recovery Centers (DRC) via any known method including postal service, deposit boxes or any occasional courier collecting service. In more general case, any method of delivery can be envisaged to be accelerated with drones, as they could be used for delivering goods between manufacturing site and logistic center, or performing partial delivery between logistic centers and automated distribution outlets/machines, or between different outlets, or between logistic centers and temporary locations indicated by human courier that could be traveling in limited delivery area, requesting drone deliveries on the go, literally experiencing raining goods in front of human delivery agents sometimes referred as postman. Further extending the concept, it is not possible to specify precisely which role any given drone will be executing prior or during the flight. Certain drones have dual or multiple capability like supporting remote delivery by partial delivery of components to Replenishment Outpost, providing partial delivery, replenishment, documenting suspected drone loss site. This is important as few rare drones that might be equipped with cameras or other sensors might on one occasion serve as heavy delivery drone, or for any other task, without any reconfiguration, or after partial in-flight reconfiguration, of performing several unrelated sub-tasks within proposed non-deterministic navigation system performing partial tasks of which only a fraction could be devoted to actual commercial drone delivery. As a result it is improper to see any application described in this document as focused on commercial delivery, but as understanding commercial delivery as
Revising the Airspace Model for commercial drone integration
Page 46
part of broader spectrum of activities of more general organization. All last-mile commercial delivery activities could be statistically performed as side effect of mentioned more general system – which in itself is not a novel concept. At the same time in extreme case it would be impossible to tell which drone within proposed system is a delivery drone, at which moment, for how long and which one is performing supporting or completely unrelated roles like inspection of emergency landing sites. It is also possible to imagine existence of pre-defined aerial packages extending into Drone-Bundle or Air-Bundle concept, which means pre-packaged sets for specific activities, like Cat&DogBundle, GrillNow!, GoneFishinEssentials, AirTravelerSupport, WildernessSet, HotAir or similar ideas not to be explained in detail because of possible underage readers. Existence of predefined bundled packages per drone type would allow shorter pre-flight preparations, knowing their exact size but also weight distribution, detailed vibration or environmental requirements and others. Those packages could also be made available in normal retail, increasing stock to comfortable levels. Such bundled drone-ready packages are significantly more efficient to handle than for example a box of shoes, which are worst-case moving payload and should not be carried without repackaging unless carrying vehicle is very large. An example would be an airplane for transporting goats, crocodiles and two wolves or other animals that are free to wander around. Because of expected range of possible Center of Gravity variations, such airplane should be larger, yet aerodynamically less clean control surfaces would need to provide more force making entire setup less safe and significantly less cost-efficient. Anybody using pre-weighted or designated packaging is going to achieve more efficient delivery drone fleet than he who uses fully modular approach with wide variation of payloads. Mentioned Gravity Measurement Post can also serve as Crawler device which could be also a separate entity, for moving a drone from Vehicle Adjustment Building (VAB) to the launch pad. Launch pads because of requirement of launching far from human beings or immediate vicinity of vehicles might form a colony of small hut-like buildings or containers around logistic centers, or any number of launch pads of any shape and design can be placed on the roofs or side walls of logistic centers. Such transport device or a robot can be replaced by a simple board and being transported on conveyor belt, rails, slides, or hanging under a rope/line system. Important aspect of fleet management is that a device can become a part of the traffic as soon as becomes self-aware of its actual configuration, which is finalized of COG measurement post. This allows it to start publishing his own capability grid like range/time/payload combo and start downloading meteo reports or other data. Other possibilities include using elevators extending from inside logistic center through the roof, or using existing or dedicated external elevators for transporting Crawlers to building roof. Those external elevators would be likely open or semi open, or at leas constructed from non-metal elements in order to equalize drone before takeoff with weather conditions as well as allow better GPS lock before takeoff. Preferred takeoff for drones would be towards the middle of the roof, and guarded by automated vision of laser proximity detectors for confirming successful launch. Unsuccessful launch may or may not require immediate drone recovery as current stock levels might allow retry with another drone, giving more time for proper failure assessment, without requiring constant presence of human supervisor on the roof. It is important to note that delivery time savings may result form bypassing existing packing or sorting facilities, either by using dedicated VAB processing route by using Integrated Payload concept.. Further generalization of GMP is a preflight measurement post allowing determining moments of inertia of payload, any of subcomponents or entire airframe, for further refining drone control loops, widening control safety margins.
Revising the Airspace Model for commercial drone integration
Page 47
Entirety of proposed solutions is therefore a blended mix of predominantly Partially Dismantled Vehicle Operations PDVO based on subtractive model, Variable Geometry Vehicle Operations VGVO, Modular Air Delivery MAD, Made-On-Demand Vehicle Operations MODVO. All of those processes of flight preparation can be supported within proposed navigation and scheduling framework. Important element of proposed system, due its potential complexity with growing number of vehicle versions, is a drone itself capable of communicating with local setup system, sharing its own pre-flight preparation capabilities. From this point of view, only generic decision making is left for ground management systems or consoles, while majority or entirety of logic that might be highly innovative and specific only for given flying device design, can be proposed and calculated by drone or generalized automated transport participant by itself. Indeed, a drone itself can be interactive system hosting its own web server or mechanical user interface displaying its preflight instruction using standard web browser, built-in display, sound messages, presenting as serial port terminal, via text messages, SMS or other cellphone communication methods or any other method. That way, each generation of drones can have extended capability that can be immediately accessible to any remote logistic centre on the world. Further generalization allows one or more vehicles per launcher creating Multiple Launch Robotic System (MLRS) or one vehicle per multiple launch pads. Each preflight transport entity can have one or more MLRS, and GMS itself could be a self-propelled drone transport device, or could be used merely as temporary container of those. Such launcher systems can be pre-loaded with anticipated payload following prediction and statistics of deliveries. Such philosophy is opposite to other methods, usually expecting operation based strictly on individual goods order. Within proposed framework, at least part of the fleet is pre-fitted with goods or other mission-related payload. This again shows superiority of capability grid, which in practice might contain average number of landing places for specific unitary device without caring about granularity and plurality of possibilities as seen from longer time or distance perspective, but this data is further refined on higher resolution grid defining specific launch pads, all of the above retaining constant analysis time from point of view of a vehicle’s flight management computer. One can also include statistical properties broadcasting for all traffic participants increased landing capabilities for given area by decreasing occupancy factors of specific launch pads for example from 100% to 90% if they host a drone expected to statistically takeoff soon, or nearing its end of duty markers. This system allows smooth integration of future predictions into a grid-based launch pad capability management. Capability of landing a single vehicle on two launch pads is advantageous to both physical infrastructure management and logical analysis of capabilities. We can imagine a structure that is able to physically hold a vehicle and recharge it (like a lamp post) and auxiliary pipe, rope, cable, conveyor (either vertical or horizontal) that is parallel to this vertical structure, can be optionally mounted and is relatively easily detachable element (without modifying internal structure of original installation), we can have dual-capability set of neighboring (two or more) launch pads that may or may not be used in parallel, extending the concept of landing, occupancy and resource management. Proposed averaged, statistical and nondeterministic view of airspace properties and logistic endpoints is crucial when we consider endless possibilities of takeoffs and landings known to aviation industry that could be applied to automated commercial traffic, either logistic or not. Those include horizontal takeoff, takeoff where parts are detaching (undercarriage, JATO, unnecessary sensors as in SDVO philosophy), launcher takeoffs (actuated with rubber, springs, medieval pulley-weight and cable arrangements, piston-actuated etc), semihorizontal takeoffs with sloped runway zone (ski-jump, gravitational slide downhill, gravitational takeoff ending with a ski-jump), vertical drop from a wall or other structure, medieval ballistic machines like trebuchet, crossbow and similar inventions, vertical takeoff, using roofs of logistic centers, walls of said
Revising the Airspace Model for commercial drone integration
Page 48
buildings, using corrugated, uneven (covered with fabrics, net), sloped landing surfaces, rope/belt based landing where rope, conveyor belt, rocket-assisted cable-driven other takeoff rails, steam, nuclear-electric, or otherwise powered linear tow–cable systems, a systems where a rope, a belt or a net can be horizontal, vertical, inclined, fixed, rigid, elastic, pre-tensioned with elastic materials like springs from one, both or more sides, free to stretch but limited by energy-dissipating set of pulleys, where rope/cable/wire not necessary serving as terminal landing point but as energy dissipating devices or flight path directing devices. Physical places corresponding to said activities might be truck trailers either parked or on the move, logistic center elements itself like designated gate docks covered with elastic sheet of plastic or other curtain material, billboards, old red massive public telephone posts, public networking posts, certain donut-shaped rotating logos or any other structures typical for advertising receiving commercial network or brand, even standardized public toilets, gas stations, fixed advertisement installations in form of round pillars, public transportation components including vehicles or roofs of zones designated for passengers, cash dispensers, existing mailboxes or popular delivery operator’s automated package dispensers. On more ironic side, proposed solution has nothing to do with mindless swarm of drones living on their own and doing random things. For example, within proposed system, it is not possible for a drone to decide to kill a customer in order to increase global customer satisfaction by reducing uneasy feeling of overpopulation just because a drone became numerically ‘hot’ thus likely to select random outcomes, alone and low on battery. One might speculate that a rebel drone after randomly selecting several path+actions has evaluated that mentioned ridiculous action was more energetically favorable to global income and human safety than another probability of a meteor obliterating entire earth and therefore less invasive solution has been selected. The truth is that building blocks of seemingly complex formulation are in fact based on 200-years old mathematics and very simple deterministically prescribed building blocks of actions. Those elementary actions are stiff and well defined; their implementation is very laborious in itself (and barely working if not correctly done, but easier to test as those are short action sequences). Granularity of those actions is also well defined; those do not include incomplete and unsafe to analyze actions like jokingly closing delivery bay or turning a propeller once hoping somebody gets hit. In order to make drones work at all, their logic must be very precisely implemented using common, existing and publicly known since several years drone position and navigation control algorithms. Except that this time, small chunks of actions are being used instead of complex flight plan, and hashed mix of those path+action building blocks is being used by entire navigation system to work cooperatively. There is no available processing power nor logic to consider any truly hidden artificial intelligence inside, even when observing that functionality of entire drone swarm, rapid decision making and complexity of factors taken smoothly into consideration is exceeding any decision making abilities of all human air traffic controllers, handicapped by their simultaneous display analysis abilities, incomplete data, slow user interfaces and fatigue. But at the very end point of technology evolution, it could be that any function evaluating most preferred outcomes for any given drone would collect a huge recorded database of small navigation decisions. Proper technology that is able to interpolate data in generalized sense by learning huge number of inputs and outputs in a compact computer code is called neural network. This is a computational structure that could generalize data in sense of filling gaps, but because of its relative efficiency could make autonomously navigating devices more compact and less expensive. At this point, while possible outcomes are still bound by navigation building blocks, traditional methods of describing workflow or decision flow would completely cease to make sense even from statistical point of view, extending simply beyond the scope of language used in any patent application, but rather described by coefficient in relatively simple neural net, being basically a computational
Revising the Airspace Model for commercial drone integration
Page 49
data structure that increases its income and customer satisfaction with very narrow set of available choices, while all of those choices/chunks would be verifiable by authorities or even general public. One could incorrectly imagine as it would be dangerous to introduce small probability that under very unusual set of conditions a drone can decide to return home by violating any forbidden airspace. First, such actions like landing in forbidden military airfield somewhere in unstable country are normally considered by any air traffic human pilot in absence of other options. Secondly, numerical weights defining factors building up traffic system energy are depending on traffic participant class. Each unmanned drone will be reasonably preset to be minimizing global drone network energy rather by deliberately jumping in a volcano (probably defined at given point of time in capability grid as low population, null retrieval chance, no drinkable water reserves, no flammable forest left, null third party damage risk, within remaining battery range, just in front, unfortunately away from profitable delivery areas) rather than flying stubbornly to nearest city over active airport with a pack of toothbrushes. Existence of alternative histories/actions to be selected with non-zero probability doesn’t mean they will ever be observed. For example, if there is only drone left, only one customer and object, a drone after every single wind gust is rethinking if it is not better to drop the idea of satisfying some payload per surface delivery factor far away defined using imprecise indication of destination position+velocity combo on some abstract delivery capability grid, and just return home. For normal flight conditions, such probabilities are easily set for such systems as equally probable as meteor striking Earth next Sunday ending all civilization. This is possible, but so unlikely it is not taken seriously for planning anybody’s action for the weekend. Additionally, such evaluation of actions millions times less favorable than other actions preferred by diminished energy of traffic system are numerically rejected due to precision of calculation for most embedded system that are and will be driving delivery drones. On the other hand, it is very important to deliberately introduce randomness to proposed traffic navigation. Let’s imagine that after several years of operation, a newly constructed windmill farm has reoriented most safety-aware system in existence by scheduling huge amount of flights over innocent, forested passage somewhere in a small valley. While classified as suspicious situation by supervisors of automated navigation system, until proper adjustment are made, we could witness a dozen drones over this seemingly sleepy area per hour, where unbeknownst to local population a large bear is sleeping. We don’t want to test his reactions when he wakes up. Introducing random element of sub-path navigation will naturally smooth drone traffic density and make them less intrusive in cases when we believed they are completely harmless, even when they will be operating at altitudes when ground animals will not even hear them – yet we don’t want to find any exceptions. If strictly deterministic traffic management system should be enforced by authorities for some irrational historical reason, one can set a temperature related to willingness to change flight plan of any traffic participant to zero. Even in such case, because amount of factors merged neatly into area preference/avoidance formulas by proposed set of dynamic capability grids, it still becomes impractical to ask for a workflow of all actions and the outcomes for the same reason it is worthless to ask for a complete workflow of a holiday travel. It is not meant that such travel would be dangerous, if we define safe building blocks of path and actions and keep them fixed. Still, freezing a list of actions for entire holiday as if it was planned by some sort of virtual Central Travel Agency might lead to seriously suboptimal results. What we see in mentioned patents or applications are precisely several holiday trips, some are missing the luggage, the others are missing weather conditions, ignoring probability of getting lost or lack of network availability (the last one is severe). In practical degraded environment devoid of positioning system and total loss of data link, patented solutions at
Revising the Airspace Model for commercial drone integration
Page 50
best enter rarely tested obscure set of laboriously crafted single-plan emergency routines, while proposed in this document decentralized weighted stochastic decision making model would offer progressive transition of drones into degraded functionality, choosing outcomes avoiding human interaction and progressively less probable deliveries, ultimately heading drones to position when they would run out of options and crash/land over safest identifiable area, taking into account balanced environmental impact information stored in computationally efficient capability/area access grids. Presented ideas are an example of what educated production line worker can invent during his free time activities and are good starting point for patent examiners. Packing several ideas into one single mind flow is intentional for protecting patent bureau public workers from laborious task of citing multiple documents for any single attempt of takeover of most banal ideas by companies searching for monopoly over not making things operational, even if those companies are experiencing only temporary troubles due to misguided global policies excluding citizens of certain nationalities from taking best paid leadership roles, even when such policies are strictly against economic interest of vast majority of its shareholders.
Revising the Airspace Model for commercial drone integration
Page 51
References This document is available in digital form at drive.google.com/open?id=0B9-O1_5eWoN9WjAzREZIX2JtMFk [REV]
Revising the Airspace Model for the Safe Integration of sUAS www.smalluavcoalition.org/wp-content/uploads/2015/07/Amazon_Revising-the-AirspaceModel-for-the-Safe-Integration-of-sUAS.pdf utm.arc.nasa.gov/docs/Amazon_Revising%20the%20Airspace%20Model%20for%20the%20Safe% 20Integration%20of%20sUAS%5B6%5D.pdf
[AMZNREV] Amazon revenue estimate www.statista.com/statistics/273963/quarterly-revenue-of-amazoncom [NPASS]
Passenger number estimate www.rita.dot.gov/bts/press_releases/bts018_16
[NBIRDS]
Bird number estimate www.motherjones.com/kevin-drum/2011/03/how-many-birds
[BARODRIFT] Estimating altitude barometer drift for sUAS diydrones.com/profiles/blogs/terrain-following-using-flexipilot-milestone-flight-report [DOF]
FAA Digital Obstacle File www.faa.gov/air_traffic/flight_info/aeronav/digital_products/dof/
[SUPM]
Secrets of UAV photomapping s3.amazonaws.com/DroneMapper_US/documentation/pteryx-mapping-secrets.pdf
[FARBI]
FAR PART 33 SECTION 76 Bird Ingestion Standards
[BUILDRES] EN 12600, EN 14019, EN 13049 [TLIN]
Environmental, Health, and Safety Guidelines for Electric Power Transmission and Distribution http://www.ifc.org/wps/wcm/connect/66b56e00488657eeb36af36a6515bb18/Final%2B%2BElectric%2BTransmission%2Band%2BDistribution.pdf?MOD=AJPERES&id=132316215484 7
[RKE]
Explaining en.wikipedia.org/wiki/Rotational_energy
[DALT]
Explaining en.wikipedia.org/wiki/Density_altitude
[INERT]
Explaining en.wikipedia.org/wiki/List_of_moments_of_inertia
[APC]
www.apcprop.com/product_p/lp27013e.htm
[PROPSAFE] Propeller Safety application www.google.com/patents/US20160039529 [ESYU]
EasyUAV parameters drive.google.com/open?id=0B9-O1_5eWoN9N0ZGNDNhaXlDdEE
[SWING]
SenseFly Swinglet parameters www.sensefly.com/fileadmin/user_upload/sensefly/images/BROCHURE-swingletCAM.pdf
[GATE]
Trimble Gatewing parameters
[EIATL]
EIA databases
geotronics.es/files/products/192/gatewing_product_leaflet_120924_rgb.pdf www.eia.gov/maps/layer_info-m.php [PPOST]
Pigeon Post presentation www.youtube.com/watch?v=3boZtKRseSA
Revising the Airspace Model for commercial drone integration
Page 52
