Session B2a: Advancements in Navigation Algorithms

Utilization of 3DMAP and Cloud/Edge to Enhance RTK-GNSS in Challenging Environments
Kaito Kobayashi, Nobuaki Kubo, Fredeluces Ellarizza, Tokyo University of Marine Science and Technology
Location: Beacon B

Utilization of 3DMAP and Cloud/Edge to Enhance RTK-GNSS in Challenging Environments

Real-Time Kinematic positioning (RTK), a precise positioning method using a Global Navigation Satellite System (GNSS), is routinely employed in intelligent transportation systems (ITS) and various other applications that demand precise positioning. Precise point positioning with Real-Time Kinematic positioning (PPP-RTK), is another promising positioning technique, similar to RTK. In open areas, ambiguities in the carrier phase can be reliably resolved, allowing RTK to achieve centimeter-level accuracy. However, in urban areas, issues, such as multipath and non-line-of-sight (NLOS) conditions occur due to high-rise buildings. These conditions prevent reliable resolution for carrier phase ambiguities, degrading, positioning accuracy to meters, or even tens of meters. Furthermore, GNSS signal outages often occur owing to signal attenuation. To resolve this, inertial measurement units (IMU) and speed sensors are integrated with GNSS to provide continuous positioning services in urban canyons.
We recently demonstrated the predictive performance of RTK-GNSS coupled with machine learning using 3DMap at ION GNSS+ 2024. The accuracy of RTK FIX predictions is approximately 70-80%. The 3D model selected was the PLATEU Project level of detail 2 (LOD2) model provided by the Japanese Ministry of Land, Infrastructure, Transport, and Tourism. The availability of a cloud server that enables real-time selection of satellites with good signal reception will permit improvement in the reliability and availability of precise real-time positioning. This technology improves GNSS positioning performance by switching between satellites to maintain high quality positioning capabilities. More concretely, we send our current approximate position to a cloud server that provides the satellite with the LOS signal. The approximate position is deduced using a car navigation device. Modern car navigation systems have inertial measurement units (IMUs) and speed sensors, guaranteeing an accuracy of a few meters, even in urban areas. This location with this accuracy is sent to the cloud server, which then estimates the satellite information that should be used by the user at that location and returns that information to the user. The user selects a satellite, based on this information and performs precise positioning. This process introduces a small time delay, that must be considered.
In this study, we demonstrate that the proposed method improves actual RTK-GNSS performance. We obtained and compared two datasets spanning approximately 1 h from a dense urban area in Tokyo without satellite selection by 3DMAP (PLATEU project) and with satellite selection by 3DMAP (PLATEU project). In a first experiment, car positions were based on precise post-processed positioning using Applanix POSLVX. The estimated positional accuracy is within 10 cm. Based on these positions, satellites with LOS signals were selected using 3DMAP, and only those satellites were used for RTK-GNSS. The results show a clear improvement relative to the dataset obtained without 3DMAP. The fix rate of each RTK-GNSS was improved by 10-15 % and DGNSS accuracy was dramatically improved. In a second experiment, car positions were based on car navigation output. The estimated accuracy of their positions was approximately 3 m for horizontal RMS. Based on these positions, satellites with LOS signals were selected using 3DMAP, and only those satellites were used for the RTK-GNSS. The results show a clear improvement relative to the data obtained without 3DMAP. The fix rate improved, similar to previous results, and DGNSS accuracy was also improved. We also investigated the performance of using the results estimated at the user's location 5 or 10 s prior, considering the time required for actual transmission to the cloud side and the analysis as the delay time. These results also show a clear improvement. Although we have used single criteria to select satellites with good quality if the received signal is LOS or NLOS. In addition, we are developing AI based satellite selection. If we could achieve better results more, we would like to present about AI based satellite selection at the ITM conference.