ABSTRACT: Three types of portable units with GNSS raw data recording capability are assessed to determine static and kinematic position accuracy

Transcription

1 ABSTRACT: Three types of portable units with GNSS raw data recording capability are assessed to determine static and kinematic position accuracy under various environments using alternatively their internal antennas and externa geodetic quality antennas. The specific units considered are single frequency and consist of the Android Huawei P10 [running on Android version 8] that outputs GPS, GLONASS and Galileo code, Doppler and carrier phase data, a custom version of the new Garmin Fenix 5 Plus sport watch that can output GPS L1 and Galileo E1 code, Doppler and carrier phase data, and the Garmin GPSMap66 released in early September The latter also output the above raw data in RINEX for either GPS, GPS + Galileo or GPS + GLONASS. The two Garmin products are all-weather robust units; the Huawei P10 is not and is sensitive to temperature change under field conditions. All above units, which are equipped with high sensitivity chipsets capable of operating in highly attenuated signal conditions, output measurements every second. The evaluation focusses on GPS and Galileo data. Static and kinematic test results obtained under a wide range of field conditions are reported. PPP and differential GNSS methods are used to assess absolute position accuracy in ITRF coordinates, which is sufficiently close to the GPS and Galileo reference frames for the current purpose. Static tests ranging from high to low multipath environments are first reported. Under ideal horizon-to-horizon line-of-sight low multipath conditions, measurements on all units are found to be sufficiently stable and accurate to provide single epoch, bias free coordinate accuracy of a few metres. Carrier phase measurements, when available, further improve performance, although carrier phase stability is a frequent issue. Static PPP coordinate accuracy is a function of measurement time interval. Kinematic accuracy performance in pedestrian and vehicular mode is a function of antenna shading, signal shading due to man-made and natural obstructions surrounding the user. Hence, performance varies widely. In all static and kinematic cases, the use of external geodetic antennas significantly reduces measurement noise and multipath, and significantly improves the quality of carrier phase data, confirming that the GNSS antenna in portable units is a major limitation. 1

2 2

3 The GPSMap66 was released in early September Raw data in RINEX is available as a standard option on all units. The Huawei P10, like numerous phones, uses a planar inverted-f antenna (PIFA), a disadvantage compared to the helix antenna used on GPSMap 66. The PIFA is a spherical antenna, hence a 3 db loss compared to hemispherical antennas designed specifically for GNSS. Spherical antennas are required to acquire and track signals in any phone position. The antenna type used in the Fenix 5 Plus watch is not known. 3

4 The above results were obtained under field conditions in a low multipath environment. As expected, the high performance Trimble R10 geodetic receiver GPS L1 and Galileo E1 code measurements are the best. The GPSMap 66 RMS values are however excellent given its low cost chip. The P10 has likely a similar quality chip but its antenna would result in higher multipath. A cursory analysis of the carrier data of the three low cost units listed above revealed noise values well below 10 mm. 4

5 ITRF2014 is the International Terrestrial Reference Frame 2014 and is compatible to the reference systems use by GPS and Galileo at the dm level or better, hence more than enough for the single point positioning comparisons reported in the sequel. RTKLib is a mature open source software used widely for differential code and carrier phase processing. 5

6 The city playground showed above is still subject to low multipath from trees and residences some 100m away. The mountain summit shown was at an elevation of 2,200m and is open sky horizon to horizon. The receivers were on the ground during the test shown. 6

7 Coach Hill Playground Results. The GPSMap 66 C/N o values are close to those of the Trimble R10, demonstrating good helix antenna performance. The Fenix 5 Plus and P10 are significantly affected by supposedly low multipath and their C/N o values are 7dB lower than those of the R10 and GPSMap 66, a significant drop that affects noise. 7

8 Coach Hill Playground single point position results. These results and other single point results shown in the remaining slides are based on code measurements only, epoch-to-epoch and no filter applied. The RMS agreement of the coordinates with derived ITRF 2014 values is excellent for the GPSMap 66 and gradually worsen for the Fenix 5 Plus and P10 units, again the effect of increasing noise and multipath. During Summer 2018, the effect of the ionosphere in the general area of Calgary was low, as measured by comparing Trimble R10 single and dual frequency positions and rarely exceeded 10cm. In differential single frequency mode [as reported in the forthcoming slides], the differential effect of the ionosphere would have been negligible. The above GPSMap 66 GPS plots show the single point and code and differential code and carrier phase results obtained with RTKLib. The reference station, equipped with another Trimble R10, was 5 km away. The RMS agreement of 24, 8 and 13 cm in latitude, longitude and height is excellent, the more so considering that RTKLib was operating in float ambiguity mode. No attempt was made to resolve the ambiguities as integer values to produce realistically achievable results under field conditions. In float mode, the code measurements and their noise and multipath contribute continuously to the solution. This is why the agreement is not at the cm level. 8

9 Coach Hill Playground results. These results were obtained at the same time and location as those described in the previous slide but using Galileo only. Galileo code and carrier measurement performances were similar to those of GPS. Position performances shown above are lower because a lower number of Galileo satellites was available and satellite geometry was poorer during in the area during Summer This will change as the four Galileo satellites launched in July 2018 start transmitting by the end of

10 Surrounding metallic objects and [not seen in the photos] nearby buildings create multipath at a comparatively high level. The coordinates obtained by the units under test were directly compared to the ITRF 2014 coordinates of the pillar used. The tested do not have ports to connect external antennas. The NovAtel geodetic antenna used as the external one was connected to the box shown in the 3 rd figure, with an open ended coaxial cable connected to the box and transmitting inside where the unit under test was placed [and surrounded by RF absorbent material]. Once the box was closed, the antenna of the unit under test was acquiring and tracking the strong signals transmitted from the outside antenna via the open ended wire. More information is provided in the following paper available on PLAN.geomatics.ucalgary.ca: Siddakatte, R., A. Broumandan and G. Lachapelle (2017) Performance evaluation of smartphone GNSS measurements with different antenna configurations. Presentation at International Navigation Conference 2018, Royal Institute de Navigation, Brighton, U.K., November, 17 pages. 10

11 These C/N o values illustrate the high multipath occurring in the environment shown in the previous slide. The R10 C/N o values are still affected by multipath but to a lesser degree than the GPSMap 66 with its own [labeled internal ] helix antenna [operating outside the box] on the lower left hand plot. When the unit is placed inside the box and track signals with the external NovAtel antennas, C/N o values [lower right hand plot] are as stable as those of the R10 unit. Again this demonstrates the importance of antennas, the more so under high multipath. 11

12 Same type of results as in previous slide but with the Garmin Fenix 5 Plus and Huawei P10 own antennas [left] and external [right] NovAtel antenna. The effect of multipath when the units are operating with their own antennas is much higher than that of the GPSMap 66 equipped with a helix antenna. The better results with the use of the external antenna are nevertheless impressive. 12

13 The plots show the GPS single point coordinates of the GPSMap 66 at the location shown in Slide 10. The high variations are due to high multipath as expected. The Huawei P10 results are specially poor, due to its antenna characteristics. 13

14 Same results as in previous slide but using Galileo satellites only. Not enough Galileo satellites could be observed with the Fenix watch and P10 to produce Galileo solutions, likely due to low C/N o, specially at lower satellite elevations 14

15 Same type of results as in Slide 13 but using an external NovAtel antenna with the three units under test. The accuracy gain with respect to the use of the units own antennas is remarkable 15

16 Same results as in Slide 14 but using Galileo measurements obtained at the time time with an external antenna. Similar performance enhancement as in the case of GPS results. 16

17 The left hand picture show two GPSMap 66 units separated by 70cm and a reference R10 in between. The P10 units, when tested, was placed behind the sliding sunroof door, hence in the centre of the Volvo XC60 roof. This vehicular test was conducted at speeds of 80 to 100 km/h. Even the R10 reference receiver experienced occasional losses of phase lock due to signal shading along the trajectory, e.g. trees as shown on the right hand picture. For all kinematic differential tests, the code and carrier phase data was processing with RTKLib. The Trimble R10 results served as reference, although themselves affected by signal attenuation and losses of phase lock due to obstruction along the trajectory. 17

18 The C/N o values of the R10 and GPSMap 66 during the vehicle test show significant drops due to signal attenuation and multipath caused by unavoidable obstructions along the road. This resulted in frequent losses of phase lock. The tests were conducted in differential mode, with a 2 nd R10 unit at the reference station located between 7 and 20 km away from the vehicle. 18

19 The C/N o values measured by the Fenix 5 Plus and P10 were significantly lower than corresponding GPSMap 66 values due to different antenna characteristics 19

20 The single point results are comparable with those obtained during static positioning tests if one takes the different multipath conditions into consideration. The agreements of the GPSMap 66 and Fenix 5 Plus in differential mode with the Trimble R10 are well within 1m RMS in each of the three coordinates. 20

21 The P10 results in this test are very good during the first part of the test. The unit then loses phase lock and does not recover. During the first part, carrier phase quality and continuity is sufficient to produce agreement with the R10 at the 1m RMS level in each coordinate, nearly as good as those of the the GPSMap 66 and Fenix 5 Plus watch. It would appear that the irrecoverable loss of phase lock is due to Android 8 limitations. Android 9, which came out after the Summer tests, is supposed to overcome these limitations [e.g. disabling of duty cycling and other issues] 21

22 These comparisons use differential positions. Since the 43 km trajectory was run twice, it was possible to compare heights for matching horizontal coordinates for each unit under test and the reference Trimble R10 unit. The R10 RMS agreement is 8 cm. If phase lock would have been maintained continuously, it would be even better. However occasional losses of phase occurred as discussed in previous slides. The GPSMap 66 RMS agreement is still better than 1m [and this includes errors in each lap]. 22

23 The upper plot show the horizontal and height differences between the differential code+carrier positions of the two GPSMap 66 units, i]. The heights of the two units on the car roof were the same; in the absence of errors, the mean height difference and RMS should be 0. The mean distance between the units should be 70cm, their fixed distance. The departures of the results from these known fixed values are reliable performance metrics of the GPSMap 66 units under realistic field conditions. 23

24 A side objective was to determine the heights of selected mountains in Kananaskis Country, west of Calgary. The 10 mountains selected ranged in height from about 2,000 to over 3,100m. The units under tests were compared to the Trimble R10 heights. PPP solutions based on 30 to 90 minutes of data [duration depended on the variable weather conditions occurring in high mountains] generally resulted in height accuracy of 1m. 24

25 Related investigations: Lachapelle, G., P. Gratton, J. Horrelt and E. Lemieux. Performance Assessment of a Los Cost Hand Held GNSS Receiver s Raw Code and Carrier Phase Data. Proceedings of 16 th World Congress of the International Association of Institutes of Navigation, 2018, Tokyo, 28 Nov - 1 Dec. Lachapelle, G., P. Gratton, J. Horrelt, E. Lemieux and A. Broumandan (2018) Evaluation of a Low Cost Hand Held Unit with GNSS Raw Data Capability and Comparison with an Android Smartphone. Submitted to Sensors, MDPI, October Siddakatte, R., A. Broumandan and G. Lachapelle. Performance evaluation of smartphone GNSS measurements with different antenna configurations. Presentation at International Navigation Conference 2017, Royal Institute de Navigation, Brighton, U.K., November