Research On Fusion Location Of Driverless Vehicle Based On ROS

With the rapid development of computer science,human's production and life are becoming more and more intelligent,and driverless vehicles play an important role in providing intelligent services to human beings.The increasing frequency of driverless vehicles in catering,express logistics,mobile travel,construction and exploration has made human activities more convenient,efficient and safe.The research of driverless vehicles is extremely important for the country's innovation ability and economic development.Real-time positioning is an indispensable important direction in the field of driverless vehicles research,which can provide strong support for the position determination and path planning of driverless vehicles.This paper is based on a driverless vehicle,combined with ROS(robot operating system),odometer,IMU(inertial navigation component)and Lidar to obtain a relatively accurate global positioning of driverless vehicles.ROS is a new standard software framework widely used in the field of driverless vehicles.It implements sensor measurement data,positioning algorithms and communication between different modules through nodes and topics.Based on the relative positioning information of the driverless vehicle provided by the odometer and the IMU,the Lidar can further realize the relatively accurate global positioning of the driverless vehicle.The main work involved in this article are as follows:(1)Firstly,the relative motion information returned by the odometer is used as the input control quantity to establish the odometer motion model of the driverless car,which will sample the pose of the car in Monte Carlo positioning.Then,based on the lidar measurement data,an observation model of the driverless vehicle is established based on the likelihood domain,and the model will be used to calculate the weight of the sampled vehicle pose.(2)Configure the firmware parameters of the VESC.These parameters are mainly important parameters required for the operation of the BLDC motor.The important parameters involved in converting the desired speed and desired angle of the driverless vehicle into the desired rotational speed of the drive motor and the desired rotational angle of the steering motor are calibrated.According to the actual speed of the drive motor and the actual rotation angle of the steering motor fed back by the VESC,the actual speed and angular velocity of the vehicle are obtained,and the odometer information of the vehicle is obtained through the track estimation.Select the model of the IMU and calibrate the parameters of the accelerometer,gyroscope and magnetometer.The rotation matrix is used to represent the attitude of the driverless car,and the rotation matrix is normalized to eliminate the numerical error.The attitude angle of the car is calculated according to the rotational kinematics,and the direction cosine matrix algorithm is used to compensate the drift of the gyroscope to obtain relatively accurate direction information.(3)The extended Kalman filter algorithm for integrating the odometer and IMU information,particle filter algorithm and KLD sampling based adaptive Monte Carlo localization algorithm are studied.Then,the fusion effect of the extended Kalman filter algorithm and the effect of particle filter estimation on discrete variable state are verified by simulation.The particle degradation of four different resampling methods is also compared.And the specific effects of each step of adaptive Monte Carlo positioning are demonstrated.(4)Firstly,a test platform was built with a car chassis,VESC,SparkFun 9DoF Razor IMU M0 and RPLIDAR-A2 single-line lidar.Then,a specific positioning scheme is designed: the VESC and the motor are used to realize the motion control of the vehicle;the extended Kalman filter algorithm is used to fuse the pose information from the odometer and IMU to eliminate the accumulated error caused by the odometer alone,thereby realizing the relative positioning of the vehicle.According to the odometer motion model and the likelihood field based lidar observation model,the adaptive Monte Carlo localization algorithm is used to realize the global positioning of the vehicle.Finally,the effectiveness of augmented Monte Carlo localization to solve the problem of car “kidnapping” and the actual positioning effect of adaptive Monte Carlo localization are verified by experiments.