基于多传感信息融合的智能车辆定位方法  

A Localization Method Based on Multi-Sensor Information Fusion for Intelligent Vehicles

作　　者：赵一兵[1] 刘昌华 郑震 郭烈[1] 马振强 韩治中 ZHAO Yibing;LIU Changhua;ZHENG Zhen;GUO Lie;MA Zhenqiang;HAN Zhizhong(School of Automotive Engineering,Dalian University of Technology,Dalian 116024,Liaoning,China)

机构地区：[1]大连理工大学汽车工程学院,辽宁大连116024

出　　处：《汽车工程学报》2021年第1期1-10,共10页Chinese Journal of Automotive Engineering

基　　金：国家自然科学基金(51975088,51975089)。

摘　　要：针对高精度定位系统中地图的重要性问题,将定位问题分为无地图定位与基于地图定位,分别对智能车辆的定位问题进行探索。对研究的智能车辆、传感器及其定位问题进行建模分析,再对该平台实施传感器校准以减小系统误差。对于无地图定位问题,利用扩展卡尔曼滤波算法将里程计与惯性测量单元(IMU)数据相融合,通过试验证明航迹推测法存在累计误差,不适用于长距离位姿估计。对于地图定位问题,采用激光传感器构建室内环境地图,根据蒙特卡罗算法(粒子滤波算法)融合里程计、IMU、激光数据信息进行室内定位试验,结果表明,基于地图的定位方法可对累计误差进行校正,在该情况下位置定位成功率可达70%以上,角度估计成功率在直线轨迹情况下高达90%,证明了定位系统中地图的重要性。Due to the importance of the map in the high-precision positioning system,localization in this paper was divided into mapless localization and map-based localization,and the intelligent vehicle positioning was explored.Initially,the intelligent vehicle platform,sensors and their positioning were modeled and analyzed and then the sensor calibration was performed on the platform to reduce the system error.Then for the mapless positioning,the extended kalman filter algorithm was used to fuse data of the odometer and the inertial measurement unit(IMU).The accumulated error was found in the experiment which indicates that the track prediction method was not applicable to the estimation of long distance pose.Finally,for map-based positioning,the laser sensor was employed to construct the indoor environment map.According to the particle filter algorithm based on Monte Carlo method,the odometer,IMU and laser data information were fused to conduct experiments of indoor positioning.The results show that the map-based localization method can correct the accumulated errors.In this case,the localization success rate can reach more than 70%,and the success rate of angle estimation can reach 90%for linear trajectory,which proves the importance of the map in the positioning system.

关 键 词：智能车辆 定位 多传感信息融合 扩展卡尔曼滤波 蒙特卡洛算法 粒子滤波算法 惯导测量单元 

分 类 号：U463.6[机械工程—车辆工程] TP212[交通运输工程—载运工具运用工程]