多传感器信息融合的自动驾驶车辆定位与速度估计  

Automatic Vehicle Location and State Estimation Based on Multi-Sensor Data Fusion

作　　者：彭文正 敖银辉[1] 黄晓涛 王鹏飞 PENG Wenzheng;AO Yinhui;HUANG Xiaotao;WANG Pengfei(School of Electromechanical Engineering,Guangdong University of Technology,Guangzhou Guangdong 510006,China)

机构地区：[1]广东工业大学机电工程学院,广东广州510006

出　　处：《传感技术学报》2020年第8期1140-1148,共9页Chinese Journal of Sensors and Actuators

摘　　要：针对大学生无人驾驶方程式(FSAC)于地图内高精定位以及速度观测的问题,设计基于多传感器信息融合的状态估计算法,并应用于自研实车平台。算法基于迭代扩展卡尔曼滤波(iEKF)进行设计,融合多类传感器,包括惯性测量单元(IMU)、转角及轮速编码器、全球卫星定位(GPS)、相机与激光雷达(Lidar)。首先,利用IMU预测车辆先验状态;然后,建立并联融合架构,对各类传感器数据进行不同的信息处理,用于更新先验状态;由于并联融合的架构,不同传感器可独立地维护车辆的状态观测。实验结果表明,所提出的算法对地图内定位、速度观测有较好的精度,且具有足够的冗余性和实时性。In order to deal with the high-precision positioning and speed problem for Formula SAE(FSAE)vehicle,a new multi-sensor data fusion estimation method based on filter is proposed and applied on real vehicle platform.The method is designed based on Iterative Extended Kalman Filter(iEKF),and combine different sensors,including the IMU,wheel speed and steering encoders,the GPS,Camera and Lidar.First,the IMU is used to predict vehicle prior state;Then,the other sensors are respectively subjected to different information extraction in parallel fusion architecture,and updated the prior state.Because of the parallel fusion architecture,different sensors may update the vehicle state vector independently.Experimental results show that the proposed method has high precision dealing with the positioning in map and speed estimation problems,and has sufficient redundancy and real-time performance.

关 键 词：自动驾驶 自主定位 状态估计 迭代扩展卡尔曼滤波 在线估计 多传感器信息融合 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]