基于激光雷达的停车场车辆定位算法

Lidar-based Localization Algorithm of Vehicle in Parking Lot

摘要:

在停车场、隧道中GPS、Wi-Fi信号受限的情况下，提出一种基于激光雷达的车辆自主定位方法。采用激光雷达SLAM（simultaneous localization and mapping）算法，通过三维激光雷达点云匹配获取车辆的估计位姿；根据图优化方法和非线性优化方法，对所有位姿进行后端调整，进而得到分辨率可控的环境信息平面栅格地图；基于蒙特卡洛方法，采用粒子滤波器进行实时车辆定位，并提出了粒子采样的一种改善方式，实现了较高精度的激光雷达自主定位。结果表明：粒子滤波器能够有效地实现车辆在停车场等无GPS环境下的定位，定位精度在10 cm之内。

Abstract:

Under the limit of car localization sensors such as GPS and Wi-Fi in parking lots and tunnels， an autonomous self-localization method of vehicle based on lidar is proposed. The lidar simultaneous localization and mapping （SLAM） algorithm is used to obtain the estimated pose of vehicle through three-dimensional lidar point cloud matching， and all poses are adjusted according to the graph optimization method and the nonlinear optimization method. Then， a planar grid map of environmental information with controllable resolution is obtained. Based on the Monte Carlo method， a particle filter is adopted for real-time vehicle localization， and an improved method of particle sampling is proposed to achieve real-time high-precision autonomous localization of vehicle. Experimental results show that the particle filter can effectively realize the localization of vehicles in parking lot and other non-GPS environments， and the localization accuracy is within 10 cm.