| 研究実績の概要 |
Autonomous driving requires lane-level positioning accuracy, in the other word, 1.5 meters accuracy. We focused on urban canyon environments such as Hitotsubashi and Shinjuku. The GNSS performance in these areas are strongly affected by multipath effects and NLOS receptions. To solve these difficulties, we used 3D building model and ray-tracing simulation to estimate the reflection effects caused by buildings. A position hypothesis positioning is proposed to apply the estimated reflection effects. We called it 3D-GNSS. This 3D-GNSS positioning achieves 5.1 meters of positioning accuracy using GPS, GLONASS and QZSS signals. An accuracy estimation method is proposed to calculate the reliability of 3D-GNSS result. By selecting the reliable result, we achieve 4.4 meters in 1-sigma positioning error. To exclude the abnormal NLOS reflection, a consistency-check that similar to RAIM FDE algorithm is developed to be used in 3D-GNSS, hence, the positioning accuracy is further improved to approximately 3 meters. In very deep urban canyon that covered by buildings and foliage, only a few GNSS signals can be received even if using high sensitivity receivers. To bridge the navigation service while the termination of GNSS signal, INS is a promising candidate. We implemented a pocket-based smartphone PDR to combine with 3D-GNSS. The result shows the proposed integration system achieves 4.3 meters positioning accuracy while the commercial GNSS receiver achieves about 39.8 meters in Shinjuku area.
|
| 今後の研究の推進方策 |
Currently, the 3D-GNSS algorithm uses only code measurements (well-known as pseudorange). The carrier phase provided by GNSS receiver has potential to provide even more accurate measurements because its resolution is in centimeter level. However, the carrier phase measurement contains a troublesome shift error, which is called carrier ambiguity. The general approach is real time kinematic (RTK), which requires a reference station (base station) to provide code/carrier corrections. We plan to develop the algorithm step by steps. Before developing the 3D map aided RTK positioning algorithm, the code-level relative positioning, namely differential GNSS (DGNSS), should be developed first. The 3D-DGNSS expects to enjoy two features of DGNSS correction, the atmospheric correction and satellite clock corrections. Moreover, the DGNSS correction also provides the clock-offset between different satellite constellation systems. This correction could benefit the previous developed consistency-check algorithm in 3D-GNSS. The second step is to implement the RTK positioning. The RTK algorithm applied double-difference measurements to retrieve the ambiguity. However, commercial GNSS receivers usually one receive measurements from single-frequency signal, which reduced the number of measurements. The insufficient number of measurements make the estimation of ambiguity even tougher. Our focus of future work will be studying how to take advantage of 3D building model and inertial navigation system to facilitate the resolution of integer ambiguity of carrier phase.
|
