Low-Cost and High-Precision Lane-Level Vehicle Positioning Method in an Urban Area Based on Traversal Optimization for MIMU Calibration Parameters

Low-cost positioning system usually consists of Real-Time Kinematic (RTK) and Miniature Inertial Measurement Unit (MIMU). For urban applications, RTK signals are often poor or lost in dense high buildings, bridges, and tunnels. In that situation, only Dead Reckoning (DR) would be used. However, DR fails to provide satisfactory accuracy limited by the parameter errors of low-cost MIMU. To address this challenge, we propose a positioning method based on traversal optimization for MIMU calibration parameters. More specifically, First, the time difference between RTK and IMU is determined by comparing accelerated velocity and angular velocity (namely motion information). For RTK, a B-spline is applied to get the motion information by calculating the derivative of the RTK position and azimuth angle. Secondly, under the criterion of minimum accumulative errors of synchronized motion information between RTK and IMU, a novel calibration method is designed based on binary tree traversal thought to obtain the optimal MIMU calibration parameters and the accurate initial navigation information. Then, using the calculated parameters and information, the DR algorithm is employed to realize high-precision lane-level positioning. We conduct extensive experiments, and the results demonstrated that the proposed method can achieve a high positioning accuracy of 0.56 m (RMSE) when RTK signals are lost for 45s, which is almost equivalent to the performance of the positioning system equipped with high-cost sensors. Our method can significantly reduce system costs and realize high accuracy for urban applications. Moreover, our method is the potential to be extended into other fields such as advanced driving assistance systems, autonomous vehicles, and mapping systems.