基于多传感器信息融合的障碍物感知追踪系统设计  

Design of Obstacle Perception and Tracking System Based on Multi-sensor Information Fusion

作　　者：吴贻杰 贾浩男 WU Yijie;JIA Haonan(School of Intelligent Science and Engineering,Harbin Engineering University,Harbin 150006,China;The 54th Research Institute of CETC,Shijiazhuang 050081,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,黑龙江哈尔滨150006 [2]中国电子科技集团公司第五十四研究所,河北石家庄050081

出　　处：《无线电工程》2023年第5期1068-1077,共10页Radio Engineering

基　　金：国家自然科学基金(62271163);新一代北斗室内外高精度无缝导航通用微终端研发(226130417A)。

摘　　要：自动驾驶系统是指采用先进的通信、计算机、网络和控制技术,对车辆实现实时、连续控制。自动驾驶研究领域中,对车身周围障碍物的感知追踪是自动驾驶车辆安全行驶的基础和前提。因此,重点研究了基于激光雷达、IMU、北斗卫星信号多源信息融合的障碍物感知与追踪系统。利用激光雷达提供的三维点云数据划分最大最小高度差栅格实现自动驾驶车辆周围的障碍物感知功能。基于激光雷达点云的障碍物感知算法中,针对特殊的悬空类型障碍物,采用高度划分的点云栅格中的点云概率密度作为悬空障碍物是否能够通行的判断依据,能够有效提升障碍物感知系统的鲁棒性。利用IMU和北斗卫星信号紧耦合方式实现自动驾驶车辆的轨迹跟踪,从而利用车辆运动状态信息结合匈牙利匹配算法实现多传感器信息融合的障碍物追踪功能。在基于多传感器信息融合的障碍物感知追踪系统中,利用IMU和北斗卫星紧耦合方式得到的转移矩阵信息去除由于车辆运动造成的激光雷达点云畸变,从而提升车辆障碍物感知追踪系统的准确度。通过实际采集得到的激光雷达、IMU及北斗卫星数据验证了提出的基于多传感器信息融合的障碍物感知追踪系统的有效性和实用性。Automatic driving system is the utilization of advanced communication,computer,network and control technology to achieve real-time and continuous control of vehicles.In the research of automatic driving,the basis and prerequisite for safe driving of automatic driving vehicles is the perception and tracking of obstacles around the vehicle body.Therefore,the obstacle detection and tracking system based on multi-source information fusion of lidar,IMU and Beidou satellite signals is primarily researched.Firstly,the 3D point cloud data provided by laser radar is used to divide the maximum and minimum height difference grid to realize the obstacle perception function around automatic driving vehicles.In the obstacle perception algorithm based on lidar point cloud,for the special suspended obstacles,the probability density of the point cloud in the highly divided point cloud grid is used to determine whether the suspended obstacles can pass,which effectively improves the robustness of obstacle perception system.Secondly,the tightly coupled IMU and Beidou satellite signals are used to realize the trajectory tracking of autonomous vehicles,and the obstacle tracking function of multi-sensor information fusion is realized by using the vehicle motion state information and Hungarian matching algorithm.In the obstacle perception and tracking system based on multi-sensor information fusion,the transfer matrix information obtained by IMU and Beidou satellite tight coupling is employed to remove the distortion of laser radar point cloud caused by vehicle movements,thus improving the accuracy of vehicle obstacle perception and tracking system.Finally,the effectiveness and practicability of the obstacle perception and tracking system based on multi-sensor information fusion are verified by the actual collected data of lidar,IMU and Beidou satellite.

关 键 词：多源信息融合 障碍物感知与追踪 悬空障碍物处理 IMU和卫星组合导航 

分 类 号：TP202[自动化与计算机技术—检测技术与自动化装置]