基于虚拟地标的行人惯性SLAM算法

针对当前仅使用低成本惯性传感器的行人导航系统航向角存在较大累积误差的问题,研究了一种只使用惯性传感器信息的虚拟地标构建方法,并通过匹配虚拟地标点对位置和航向进行误差补偿.在此基础上,结合同时定位与建图(SLAM)方法的思想,实现了仅基于惯性传感器的SLAM,提高了行人导航的精度和可靠导航时间.试验结果表明,该方法能够有效消除纯惯性行人定位系统的累积误差,在2027m2的单层建筑物中空间位置误差小于10m.

Pedestrian Inertial SLAM Algorithm Based on Virtual Landmark

To address the problem of large cumulative error in the heading angle of the current pedestrian navigation system using only low-cost inertial sensors, we investigate a virtual landmark construction method using only inertial sensor information, and compensate the error of position and heading by matching virtual landmark points, based on which, combined with the idea of Simultaneous Localization And Mapping (SLAM) method, we realize the simultaneous localization and composition based on inertial sensors only, and the accuracy and reliable navigation time of pedestrian navigation are improved. The experimental results show that the proposed method can effectively eliminate the accumulated errors of the pure inertial pedestrian positioning system, and the spatial position error is less than 10 meters in a single-story building of 2027 square meters.