Research On Strategy Of Driverless Car Based On GPS And Lidar

With the development of artificial intelligence,driverless cars are gradually coming into people's sight.As drivers are replaced by computers,it is hard to see "road rage","drunk driving","jam".As a result,the incidence of traffic accidents was reduced and traffic congestion was alleviated.This paper takes the driverless car of Chang 'an university as the experimental platform and focuses on the research of the driverless strategy based on GPS and Lidar.The main work is as follows:(1)A differential GPS platform was built with Trimble-BD982 and F2414,and we collected GPS points based on technology of Real-time kinematic(RTK).Neville interpolation method was used to deal with the GPS points sparse phenomenon caused by high speed.We put the GPS points from WGS-84 coordinate system to gauss coordinate system by gauss projection,and alleviated Projection distortion by zoning.A smooth GPS road network was obtained after fitting by NURBS algorithm,and we calculated the vehicle's deflected angle and distance relative to the preset point based on position and heading provided by BD982.(2)We analyzed the data packets of the HDL-64 E Lidar according to the UDP protocol,and converted the points from Euclidean coordinate to cartesian coordinate according to the calibration file.Then points were transformed from lidar coordinate system to vehicle coordinate system after the lidar position relative to the vehicle was calibrated.Removing the road surface based on the characteristic that obstacle will destroy the circle formed by the same laser and a line perpendicular to the obstacle is away from the center of the circle.The obstacles were clustered by means of the difference of distance and reflectivity between adjacent points and lidar by exogenous seed clustering.The interested region was selected based on the roadside detection,and the obstacles in the interested region were divided into static and dynamic.The minimum rectangular envelope algorithm was used to transform irregular obstacles into rectangles,and an appropriate expansion coefficient was selected to enlarge the capacity of obstacles in front of the vehicle.The enlarged rectangle was also equipped with GPS information after the transformation from the car coordinate system to the gaussian coordinate system,and then the path was designed to avoid obstacles.(3)By collecting the driving data,this paper establishes the fuzzy control with deflected angle and distance as input and rudder angle as output.In addition,this paper take a traditional navigation algorithm to solve the problem that fuzzy control can't turn the vehicle around because the input range is too narrow.The lower computer controls the speed with PID,and the rudder is controlled with the Angle interval.In the last,this paper carried out the driverless experiment of straight line,S-bend and obstacle avoidance in the automobile experimental field of Chang'an university.The experiment proved that the algorithm of tracking and obstacle avoidance in this paper are reliable,and the control accuracy of the upper computer and lower computer is high.