Research on Multi-sensor Fusion Localization for Autonomous-rail Rapid Tram

Abstract

Traditional positioning methods, such as GNSS (global navigation satellite system) and location base service, have large positioning error when applied in complex traffic scenarios such as like tunnels buildings and viaducts. Odometer has accumulated errors, so it is difficult to achieve accurate positioning over long distances. For the problem of large positioning error in complex driving environments, this paper proposes a long time and high precision positioning system based on LiDAR - global map matching to achieve robust and accurate state estimation. The system mainly consists of two parts: global map construction and map matching location. Firstly, absolute localization constraint of global navigation satellite system is introduced in map construction to eliminate cumulative errors of LiDAR odometer and construct a globally consistent high-precision map. Secondly, in the part of map matching location, the original global map matching and real-time dynamic map matching are combined to effectively avoid positioning degradation caused by environmental changes, which can improve the positioning robustness. Test results of the autonomous-rail rapid tram on open urban roads show that this method can provide vehicle pose information stably and reliably. The angle estimation error of LiDAR is within 0.35° and the position error is within 0.15 m during the whole operation process, which verifies the effectiveness and robustness of the proposed algorithm.

Keywords

• autonomous-rail rapid tram
• mapping
• localization
• multi-sensor fusion