Vehicle tracking trial results for SBAS in South Africa Information from SANSA The SATSA (SBAS Awareness and Training for South Africa) project was established to facilitate SBAS (satellite based augmentation system) deployment in South Africa by increasing technological navigation capacity in the region. During the course of the project, three trials were conducted to evaluate SBAS applications using a test platform to produce realistic SBAS signals for the duration of the trials. These trials focused on vehicle logistics, precision agriculture and GIS ground truthing. This article looks at the results of the vehicle tracking training trial. To facilitate the trials, MagicSBAS the SBAS testbed was installed at the SANSA Space Operations premises at Hartebeeshoek using data from South Africa s existing network of GNSS stations (TrigNet) in order to provide an EGNOS-like signal, which could then be accessed via the internet. Eclayr and MagicGemini software helped to determine the performance of the generated signal, and proved indispensable during the trials which were held earlier this year. Architecture for the trials In relation to the type of end user equipment needed for the trials, it was decided that the best performance would be obtained if a high-end SBAS receiver was used in the trials with a high quality antenna on the roof of the vehicle. However, the companies collaborating on the trials, preferred to use their standard receivers and installation configurations in order to focus on the differential performance that EGNOS delivers under their normal operations. In consequence, it was agreed to use, during the trials, the following architecture (see Fig. 1): MagicSBAS server in SANSA premises, providing in real time the SBAS navigation message both in SISNeT, and DGPS (RTCM 2.3) formats. The SISNet stream was used for monitoring and testing purposes during the realisation of the trials (testing of the SBAS correction performances in real time through MagicGemini capabilities). The DGPS stream was the one used by the end user equipment to compute the augmented SBAS solution. It is worth noting that, Fig. 1: Proposed "EGNOS" architecture for the SATSA trials. for each of the trials, a virtual reference station (VRS) location close (<10 km) to the trial scenario was configured in order to maximise the final accuracy performances. Two types of receivers were used in each one of the trials: - Standard terminal equipment used by the local companies (with minor modification in the case of Tracker to allow ingesting the RTCM corrections) - A high-end, double-frequency (L1, L2) receiver provided by SANSA with the objective to store the raw input measurements during the trials. The architecture described above facilitates two main objectives: The local company partners were able to evaluate during the trial duration the SBAS position solution with their own equipment. SANSA was able to evaluate, in offline mode, the performances in accuracy of the position solution by using the L1/L2 raw GPS measurements recorded during the trial. In order to compute final position accuracies, the true trajectory/ position of the vehicles during the trial duration needed to be known with centimetre level accuracy. This was achieved by post-processing the PositionIT Nov/Dec 2013 65
Receiver Novatel receiver Septentrio receiver Receiver capable of RTCM2.x or 3.x outputs Type connection Serial Port Serial Port Serial Port Output format Proprietary Proprietary RTCM 2.x / RTCM 3.x Minimum observables needed Observables processed by MagicGemini Messages needed SBAS messages Remarks GPS L1 (pseudorange and carrier phase) GPS L1 (pseudorange and carrier phase) GPS L1 (pseudorange and carrier phase) GPS + GLONASS L1/L2 GPS + GLONASS L1/L2 GPS + GLONASS L1/L2 OM4 #43: RANGE_ID OM4 #41 RAWEPHEM SBAS SIS messages are normally received by the receiver, so in this case where SISNeT messages are to be received, the MagicGemini will need to be modified Additional efforts are envisaged to make MagicGemini able to read SBAS messages from SISNeT Server SBF # 5889: MEASEPOCH SBF # 5889: SMARTMEASEPOCH SBF # 5889: NAVIGATION SBF # 5889: PVTCARTESIAN SBAS SIS messages are normally received by the receiver, so in this case where SISNeT messages are to be received, the MagicGemini will need to be modified Additional efforts are envisaged to make MagicGemini able to read SBAS messages from SISNeT Server RTCM 2.x RTCM 3.x 1(1) 1004(1) 18(1) 1012(1) 19(1) RTCM 3.x 1019 GPS and 1020 GLONASS ephemeris messages are also needed (they are generally available through Ntrip Casters) SBAS messages are received via SISNeT This is the recommended option as less effort is required in order to integrate the receiver within the MagicGemini monitoring tool Table 1: Minimum specifications for SATSA trial receivers. 66 PositionIT Nov/Dec 2013
Fig. 2: Trial car with antennae in the back window. L1-L2 raw measurements using the GMV tool MagicPPP [1]. MagicPPP is an internet based service that allows GNSS users to determine their position or trajectory with centimetrelevel accuracy. In light of the fact that all these applications are likely to use the EGSA OS in the future, it was agreed that GMV would optimise the algorithms to maximise accuracy of the OS for the training tool and trials. Fig. 3: Map of the second road trial. Equipment requirements With the above architecture in mind, GMV specified the minimum requirements for the SATSA trial receivers and these are shown in Table 1. Note that the first two columns refer to the specifications of high-end SBAS receivers, while the final column is the one that pertains to the operational GPS receivers in each application area. The collaborating companies were tasked to check if their standard receivers could output the GPS solutions in the appropriate format. Other requirements In the absence of an EGNOS SIS across South Africa, the trials utilised SBAS/EGNOS corrections for the region via the internet, which were accessed via the mobile communications network in South Africa. The amount of data that needed to be transferred was significant and thus required the higher bandwidth of 3G GSM as the bandwidth of GPRS would not be sufficient. Consequently the trials had to take place in areas with good 3G GSM coverage. In practice it turned Fig. 4: Horizontal and vertical position errors in GPS (left) and SBAS (right) modes. out that GSM EDGE was sufficient for the data streaming required. Furthermore, the best performances would be obtained if the trials were conducted in areas where there is an open sky environment. This is due to backscatter errors that are likely to occur in dense urban areas with many high buildings. Vehicle tracking trial Tracker was interested in testing their standard equipment versus the same with improved EGNOS accuracy. The vertical component is becoming more and more important to them, and thus it was agreed that altitude information from MagicSBAS would also be included in the trial. Tracker intially started off as a stolen vehicle recovery (SVR) company but has now moved into telematics and information services, especially for insurance companies that offer payas-you-drive or behavioural insurance products 1. They are partly owned by a UK company, Actis. Tracker currently has around 700 000 vehicles equipped with their Lojack tracking product (for vehicle recovery) and 225 065 vehicles with telematics systems onboard for pay- Footnote 1. For this they monitor cornering, harsh braking or acceleration, speeding, route and mileage, etc. PositionIT Nov/Dec 2013 67
9 m, with less than 5 m on occasions. To track cars on roads, they use maps from TomTom (TeleAtlas) and also face challenges regarding the accuracy of the map data. Total distance driven is a crucially important parameter for them, and for more accurate distance calculations they would also like to have altitude positioning. Currently the distance calculations have a 5% accuracy, and ideally they would like to reduce this to 3%. Fig. 5: North versus east error in GPS (left) and SBAS (right) modes. They also face issues of poor GPS positioning in urban environments with tall buildings (typically downtown Johannesburg). Trial preparation Fig. 6: Comparison of the vertical error in GPS and SBAS modes. For the trial preparation, Tracker had first to implement an own software to be able to receive the RTCM DGPS MagicSBAS corrections and send them to their U-Blox receiver, as the GNSS receivers they were using did not support the MagicSBAS standard Sisnet protocol for SBAS data dissemination as stated in Table 1. Before the trials Tracker checked that by using an antenna splitter to feed two U-Blox receivers, one running in GPS standalone mode and another one in DGPS mode Trial location and route Fig. 7: Real-time position obtained during the trial using U-Blox centre with a GPS standalone (left) and SBAS (right) solutions. as-you-drive insurance. For the telematics sector, Tracker partnered with an SA company, Microtronics, to develop their own tracking device which incorporates an U-Blox Neu 6 GPS chipset. They utilise Sierra wireless technologies and the MTN and Vodacom mobile networks in SA for communications. Tracker is currently working on their next generation device (still using the U-Blox Neu 6 GPS chipset) which includes a gyroscope. They are also developing a new product with Cobra Automotive Technologies (Italy), which typically supplies SVR systems for high end cars such as Porsche. Tracker is also the exclusive distributor of TomTom products in SA and supply TomTom with local traffic information and services. Operational requirements for improved positioning Their insurance clients are typically looking to reduce their risk through a reduction in car theft and tailoring premiums to better match driver behaviour. Improved positioning is more important for the insurance sector than for SVR. In particular for the behavioural insurance products, the insurance companies need more accurate location information of the cars, e.g. where there are two parallel roads with different speed limits. Tracker s current operational accuracy with the U-Blox GPS receiver is 6 to The vehicle tracking trial took place on 3 March 2013, starting at Tracker s headquarters in Cresta, Gauteng. A route of 125 km and nearly 3 hours was taken through a mixed environment (urban and rural) in the Sandton area, along a five lane motorway and with a rural section that included hilly terrain (to test the altitude parameter). The trial was carried out by Justin Marais from Tracker, Eugene Avenant and Farhad Hassim from SANSA and Javier Ostolaza González from GMV. Emiliano Spaltro from Alpha Consultants and Nina Costa from NDConsult used the visit to Tracker to interview Carel Wessels from Tracker on the potential business case for EGNOS in vehicle tracking and tracing. Vehicle and GNSS equipment One small car was used carrying two of Trackers standard U-Blox Neo 6 GPS receivers and one magnetic antenna set on the roof of the vehicle, with a splitter to feed both receivers with the same signal. One of the receivers was set into GPS standalone mode while 68 PositionIT Nov/Dec 2013
A comparison of the GPS and SBAS accuracies performances in the vertical direction (see Fig. 6). It is worth noting the better performances obtained with the SBAS solution than the conventional GPS-only solution. This is particularly remarkable in the vertical direction, where precision around 1 m is reached for the SBAS solution. It is also worth noting that, in the polar plot the SBAS error is very concentrated in a dense cloud, whereas the GPS errors have a more scattered pattern. Fig. 7 shows the real-time results obtained from the observations in the U-Blox centre software. These images have been reproduced from the real time data recorded during the first trial. The SBAS solution offered in real-time a better estimate of the position, considering the antenna was set on the left side of the roof of the vehicle. Fig. 8: Real-time comparison between GPS standalone and SBAS solution. the other one was set into MagicSBAS DGPS mode, receiving the SBAS corrections via the software developed by Tracker and showing the real-time position on a Google-Earth map visible through the U-Blox software centre. The trial architecture allowed for a real-time check of the different performances between both modes on Google Earth maps. In addition, SANSA s high-end GNSS receiver (Trimble NetR9) was also used in order to record pseudorange and phase raw data in order to evaluate the SBAS performances in post-processing mode, however, the vehicle for the demo was not set up to allow the antenna to be fixed to the car roof, so it was set up inside the car, below the rear glass, as can be seen in Fig 2. As is observed in the next section, some of the satellites were consequently not in view during the trial. The trial was completed in one day. Provision had been made for a second day if problems were encountered on the first, but this proved unnecessary. It was realised, when post processing the raw measurements stored by the SANSA s Trimble NetR9 receiver that the antenna location was not ideal to ensure correct visibility of the GPS constellation (the antenna was attached to the rear window of the car). In consequence, the input measurements were very noisy, not allowing a correct computation of the true trajectory of the vehicle. Furthermore, the interpretation of the poor results obtained would suggest a wrong idea of the final accuracy performances obtained in this type of scenarios SANSA decided to repeat the trial with its own vehicle during the first week of June. Hereafter the results correspond to the second round of the trial. Trial results In Figs. 4 to 6 the performances obtained using the receiver GPS only mode and GPS+SBAS mode throughout the dynamic road trial around Johannesburg and Pretoria can be observed. Note that these performances have been obtained by processing the RINEX file obtained from SANSA s Trimble NetR9 receiver using the GPS standalone and a SBAS solution (using MagicSBAS SBAS messages) and comparing the positions obtained with the reference ones computed using MagicPPP which offers centimetre error accuracy. The performances which can be observed for the different passes are as follows: The horizontal and vertical position errors for GPS standalone (left) and SBAS (right) solution (see Fig. 4). The north versus east error for GPS (left) and SBAS (right) solution where the dispersion of the horizontal errors within the two modes can be compared (see Fig. 5). Conclusions The results of the first road trial with MagicGemini were not very clear because the antenna of the Trimble receiver was inside the car, below the rear glass. However, after examining the video footage it was possible to see a deviation from the GPS and the SBAS solution. It was apparent that the SBAS solution presents a better solution, as the U-Blox antenna was placed on the left side of the roof of the vehicle. Results from the second trial show very good accuracy performances, in line with expectations for a SBAS solution. Having completed these trials and processed the data, the team has learnt the lessons required in order to plan and execute trials of this nature with greater insight which will yield even better application specific output that can be used to further the case for using SBAS in the non-aviation domain. The results from this trial show the application of SBAS in the non-aviation domain. SANSA needs this information in the process of further motivating for the development of SBAS in the country (even though in this case it does not justify the SoL case). The outputs of these trials are being incorporated in the planning documentation for SANSA such as the national space plan. Reference [1] www.gmv.com/en/space/magicppp Contact Eugene Avenanat, SANSA, Tel 012 334-5000, eavenant@sansa.org.za PositionIT Nov/Dec 2013 69