多传感器融合的无人车SLAM系统研究  

Research on multi-sensor fusion SLAM system for unmanned vehicles

作　　者：吴文昊 谷玉海[1,2] WU Wenhao;GU Yuhai(Key Laboratory of Measurement and Control Technology,Ministry of Education,Beijing Information Science and Technology University,Beijing 100192,China;Mechanical and Electrical Engineering,Beijing Information Science and Technology University,Beijing 100192,China)

机构地区：[1]北京信息科技大学现代测控技术教育部重点实验室,北京100192 [2]北京信息科技大学机电工程学院,北京100192

出　　处：《重庆理工大学学报(自然科学)》2025年第1期229-235,共7页Journal of Chongqing University of Technology:Natural Science

基　　金：现代测控技术教育部重点实验室开放课题资助项目(KF20222223205)。

摘　　要：为提高无人车的避障能力,使其能够在构建的地图环境中高效地进行自动定位和路径规划,提出一种多传感器融合的无人车SLAM系统。对于障碍物监测,采用激光雷达与深度相机信息融合的方法构建地图,以融合得到更精准的栅格图。搭建了履带式差速底盘运动学模型,通过融合IMU数据提高位姿估计精度;分析了贝叶斯推理方法,在决策层以该方法有效融合激光雷达与深度相机的数据;提出基于卡尔曼滤波算法动态调整权重将雷达与相机的后验概率融合,得到最终的地图栅格信息。最后,根据融合后的数据构建地图并实现自主导航的功能。通过对比实验发现,改进的多传感器融合建图算法定位精度综合提高了91.67%,实时的整体性能提升了54.46%,栅格建图完整性提升了6.59%。To improve the obstacle avoidance ability of unmanned vehicles and efficiently perform automatic positioning and path planning in the constructed map environment,we propose a multi-sensor fusion SLAM system for unmanned vehicles.Due to the blind spots and defects in the detection of a single radar,a map is built by employing the information fusion method of lidar and depth cameras for obstacle monitoring to obtain a more accurate raster map.First,a kinematic model of the tracked differential chassis is built to improve the pose estimation accuracy by fusing IMU data.After an analysis of the Bayesian inference method,it is used to effectively fuse the data of lidar and depth cameras at the decision-making level.Then,we propose to dynamically adjust the weight based on the Kalman filter algorithm so as to fuse the posterior probabilities of the radar and camera and obtain the final map grid information.Finally,a map is built and autonomous navigation is achieved based on the fused data.Our comparative experiments show the improved multi-sensor fusion mapping algorithm improves the positioning accuracy by 91.67%,the overall real-time performance by 54.46%and the raster mapping integrity by 6.59%.

关 键 词：贝叶斯算法 融合建图 激光雷达 深度相机 ROS2 

分 类 号：TP242.2[自动化与计算机技术—检测技术与自动化装置]