1

1

Accuracy of the Swift-Navigation Piksi differential GPS

For validation purposes the accuracy of the GPS-System was tested on a field in Aachen, Germany. During the whole test-time the weather conditions were fair. First, the set-up and the test execution will be described. Then the results of the test will be shown.

Set-up Two piksis are used for the test. One of them, which acts as basestation, is connected to a notebook via USB-Port. The other one is the rover and is connected to an Arduino-Leonardo with SD-Card shield. The Arduino-Leonardo filters the Swift Binary Protocol and the SD-Card shield is used as data storage. As we are interested in the relative distance between the base and the rover, the Leonardo is programmed in that way, that it is filtering only the BASELINE NED message and stores the vectors north, east and down between the rover and the base. Piksi STM Firmware Version v0.19, Piksi NAP Firmware Version v0.15 and Local Piksi Console Version 0.24.1 are used. The settings of the piksis are set to default and the baudrate is set to 115200. To establish a reliable communication between the two piksis, they are both mounted on a two meter pile. The 3DR-433 MHz radios with default supplied antennas are used for communication. In picture 1.1 the whole set-up is shown. A 0.7 mm fishing line is stretched between the two piles. Figure 1.2 shows a close up of the mounted piksis. The GPS-Antennas are mounted on a 10x10 cm aluminium sheet.

Test-execution Once the piksis are connected to their power supply, they start to generate a RTK-Solution. When they get a fixed solution, the rover is carried in a circle three times around the base, as shown in figure 1.3. The walking direction is changed every round. It is ensured, that the fishing line is always stretched while the rover is carried around the base. That way, it is later possible to visualize the accuracy of the relative distance between the rover and the base.

Results In figure 1.4 and figure 1.5 the results of both tests are shown. The difference in the vertical distance between the different rounds in figure 1.4 is caused through a not constant altitude while carrying the rover. While starting the test with a radius of 30 m the RTK-solution was in float modus and during walking it switched into fixed modus. That is probably the reason for the vertical difference in figure 1.5. Based upon fixed“ north and east vector an arithmetically averaged horizontal distance is cal” culated in post-processing. Then the deviation to this radius“ is calculated and plot in figure ” 1.6. The mean radius“ of each walk is slightly higher than the length of the fishing line. Rope ” stretching could be a reason for that. The standard deviation amounts to 129 mm for a 20 m walk

1

Accuracy of the Swift-Navigation Piksi differential GPS

2

Figure 1.1: Setup on the field

Figure 1.2: Close up of the mounted piksis

and to 94 mm for a 30 m walk. Taking a look at the set-up shows that the antenna of the base is not in the center of the circle. There is a difference of 7 cm between center and gps-antenna. This mistake in the set-up leads to differences in the calculation of a mean radius and hence to a bigger standard deviation. To reduce some error sources like rope stretching, not centred base-antenna or not fixed altitude of the carried antenna, a third test set-up is build up. It is shown in figure 1.7. This test is a static measurement with an exactly known distance of 2.15 m between the antennas of rover and

1

Accuracy of the Swift-Navigation Piksi differential GPS

3

Figure 1.3: Test-execution schematic

base. The horizontal distance calculated based upon north and east vector is shown on the left side of figure 1.8. The distribution plotted on the right side is normally distributed and has a standard deviation of 10.2 mm. The range of 2 x σ contains 95.5% of all values. That leads, for the static measurement, to a horizontal accuracy of ± 2 cm for the Swift Navigation differential GPS-System. 2.5

×104

2 500

1.5 1

East [mm]

Down [mm]

0 -500 -1000

0.5 0 -0.5 -1

-1500 4

-1.5 2

×104

4 2

0

0

-2

East [mm]

-4

-4

-2 ×10

-2

4

-2.5 -3

North [mm]

-2

-1

0

1

2

3 ×104

North [mm]

Figure 1.4: Results for 20 m diameter walk

4

×104

3 6000 2

2000

East [mm]

Down [mm]

4000

0 -2000

0 -1

-4000 4

-2 2

×104

1

4 2

0

0

-2

East [mm]

-2 -4

-4

North [mm]

×104

-3 -4 -4

-3

-2

-1

0

North [mm]

Figure 1.5: Results for 30 m diameter walk

1

2

3

4 ×104

Accuracy of the Swift-Navigation Piksi differential GPS

3.5

4

Density function; Mean= 20248.351mm and Sig=129.1357mm

×10-3

3

2.5

Density

2

1.5

1

0.5

0 1.99

2

2.01

2.02

2.03

2.04

2.05 ×104

Radius [mm]

(a) Distribution for 20 m walk

4.5

Density function; Mean= 30406.1719mm and Sig=94.0778mm

×10-3

4

3.5

3

Density

1

2.5

2

1.5

1

0.5

0 3.015

3.02

3.025

3.03

3.035

3.04

3.045

3.05

3.055

Radius [mm]

(b) Distribution for 30 m walk Figure 1.6: Normal distribution of the horizontal distance between rover and base

3.06 ×104

Accuracy of the Swift-Navigation Piksi differential GPS

5

Figure 1.7: Set up static measurement

0.04

Density function; Mean= 2150.7495mm and Sig=10.2026mm

0.035 80 0.03 60 0.025

Density

Down [mm]

1

40 20

0.02 0.015

0 450

0.01 2160

400

2140 2100

East [mm]

0.005

2120

350 300

2080

North [mm]

0 2110

2120

2130

2140

2150

Distance [mm]

Figure 1.8: Results for 2.15 m static measurement

2160

2170

2180