激光雷达/IMU/车辆运动学约束紧耦合SLAM算法
Tightly-coupled SLAM algorithm integrating LiDAR/IMU/vehicle kinematic constraints
杨秀建 1颜绍祥 1黄甲龙1
作者信息
- 1. 昆明理工大学 交通工程学院,昆明 650500
- 折叠
摘要
针对自动驾驶车辆在全球导航卫星系统(GNSS)信号不足场景下的定位需求,提出了一种激光雷达/惯性测量装置(IMU)/车辆运动学约束紧耦合的同时定位与地图构建(SLAM)算法.首先,基于 IMU 角速度、车辆后轴轮速和前轮转角构建车辆运动学约束,将车辆运动的位移和姿态信息解耦,构建位移和姿态约束以提高优化结果的准确性;然后,根据点云特征点数量和车辆转向角度引入自适应调整系数,实时调节车辆运动学约束的权重.最后,基于 IMU 角速度和车辆后轴轮速构建里程计模型,为后端紧耦合优化提供精准的初始值,避免陷入局部最优.不同道路场景下的测试结果表明,所提算法与 LeGO_LOAM和 LIO_SAM 算法相比,平均平面定位精度分别提高了 32%和 29%,为自动驾驶车辆提供了一种GNSS信号不足情况下的短时高精度定位解决方案.
Abstract
Targeting the positioning requirements of autonomous vehicles in scenarios with insufficient global navigation satellite system(GNSS)signals,a tightly-coupled simultaneous localization and mapping(SLAM)algorithm that integrates LiDAR,inertial measurement unit(IMU)and vehicle kinematic constraints is proposed.First,vehicle kinematic constraints are established with the angular rate of IMU,rear wheel speed,and front wheel angle.By decoupling the displacement and orientation information of the vehicle motion,the constraints for displacement and orientation are formulated separately to enhance the accuracy of optimization.Then,adaptive weights are introduced based on the number of feature points and the steering angle to dynamically adjust the weight of the vehicle kinematic constraints in real time.Finally,an odometer is constructed based on the angular rate of IMU and rear wheel speed,providing precise initial values for the back-end tightly-coupled optimization and effectively preventing falling into local optima.Results from tests conducted in various road scenarios demonstrate compared with the algorithms of LeGO_LOAM and LIO_SAM,the average planar positioning accuracy of the proposed algorithm is improved by 32%and 29%respectively,which provides a short-term high-precision positioning solution for autonomous vehicles in situations where GNSS signals are insufficient.
关键词
自动驾驶/同时定位与地图构建/多传感器融合/车辆运动学Key words
autonomous driving/simultaneous localization and mapping/multi-sensor fusion/vehicle kinematics引用本文复制引用
出版年
2024
NETL
NSTL
万方数据