Direct Estimation of Multipath in a Deep Urban Area using Multi-GNSS Carrier Phase Variation and Previous Position

Abstract

Accurate positioning in unban canyon has been a major challenge in the GNSS field for a long time. As the use of satellite navigation system of smartphones and car navigation systems become more usual in populated area and cities, the need for improving the GNSS position accuracy is continuously increasing. Multi-path and NLOS (Non-Line-Of-Sight) signal are the major cause of the poor accuracy in urban area. Recently various studies such as multi-constellation GNSS, NLOS measurement rejection, Ray tracing, Shadow matching and weighting matrix based on the signal strength have been introduced to reduce multipath effect. This paper introduces efficient and practical methods to estimate and eliminate the multi-path for both LOS and NLOS signal, and then shows this technique is very effectively utilized for automobiles in a deep urban city like Teheranro, Gangnam-gu, Seoul. In this paper, previous position information and CMC (code minus carrier phase) variation are used to estimate values of multipath error of both LOS and NLOS. CMC measurement consists of ionospheric delay, multipath, noise of observables and carrier phase ambiguity, since all the geometry terms like distance between a receiver and a satellite, tropospheric delay, ephemeris and clock errors are common to the code and carrier measurements. If we combine L1 and L2 measurements to generate ionosphere-free measurements, variation of CMC is almost same as that of code multipath, since code multipath is far larger than phase multipath and noise terms. Assuming that there is no cycle-slip between two specific epochs and the previous position is valid, the code multipath can be continuously and directly estimated based on this CMC variation regardless of the satellite in LOS or NLOS direction. To apply the suggested algorithm in deep urban, successive valid measurements are essential. The target area, Teheran-ro is a notorious street for the low visibility of the navigation satellite. In this area, only 2.4 GPS satellites are visible on average and users can calculate its own GPS position during only 20% of the day, thus multi-constellation is required for the GNSS position. However, the minimum number of visible satellites getting more when adding another constellation to GPS because of the inter- system bias. Even though the average number of the visible satellites is increased to 5.25 after adding 5 GNSS systems, i.e. QZSS, GLONASS, Beidou, Galileo and IRNSS to GPS, the position availability in Teheran-ro due to the inter-system bias is 70% a day and its error is occasionally up to 250m. To cope with the multi GNSS bias, we set the initial multipath value of other constellations than GPS to include the inter system bias based on the calculated position information. Based on this strategy, a rover’s position can be solved with only 4 visible multi GNSS satellites. To prove the suggested algorithm and to show its performance in an urban canyon, several static and dynamic tests were conducted. A test vehicle that included a Low-cost dual-frequency GNSS receiver, Swift Nav Piksi Multi was used to drive along the Teheran-ro street for one hour, and Novatel SPAN provided reference trajectories. 1 hour driving test results shows that the suggested algorithm can provide the rover’s real-time results for 94.5% of the session, while the receiver calculates its position for only 86.6%. The RMS value of the horizontal error is reduced from 8.7m to 2.96m by applying the suggested algorithm to the receiver and 95% of horizontal error was also improved by 70%, from 17.99m to 5.22m. Most impressive thing is that the position error of the algorithm is bounded to 6.14m, while the maximum error of the receiver is 58.9m.