Design And Implementation Of Multi-Lidar Coupling System For Autonomous Driving

In recent years,the rapid development of autonomous driving has become a hot topic in global research.Key technologies for autonomous driving include perception,decision,planning and control,where perception is the foundation of decision,planning and control.The perception module relies on a variety of sensors to sense the driving situation around the car,so the quality of the sensor will directly affect the perception,which further affects the safety of the driverless car.Lidar is one of the most commonly used sensors in the driverless field.Lidar has a long detection range,high precision and strong anti-interference ability.The low-configuration lidar is cheap,but the resolution is low and imaging performance of the point cloud is poor,which is not conducive to algorithm analysis;The hign-configuration lidar has a higher resolution and imaging performance of the point cloud is better,but the price is expensive and the lead time is long.Therefore,there is an urgent need for a lidar system with high resolution,low price and short lead time in the driverless field.In respect of the issues above,this paper proposes a method of coupling multiple low-configuration lidars to construct a multilidar coupling system,which achieves the goal of obtaining high resolution point clouds at a lower cost and improves the sensing accuracy effectively.This paper relies on the Robot Research Group of Jilin University and has completed the following tasks:1)To solve the problems like the high cost of high-configuration lidar,the low resolution of low-configuration lidar in the driverless field and so on,this paper designs a new multi-lidar coupling structure based on the existing four-lidar coupling scheme.The laser lines of multiple low-configuration lidars can be evenly distributed in space with the new structure,which improves imaging performance of the point cloud.2)To figure out the calibration problem of multiple low-configuration lidars,we designed a new calibration method for low-configuration lidar based on improved genetic algorithm.The method can solve the external parameters between the low-configuration lidars to merge the point clouds of the plurality of low-configuration lidars into the same coordinate system where the resolution of the point cloud improves.3)In order to verify the practicability of the multi-lidar coupling system,this paper compare the system with a higher cost lidar--velodyne HDL32 E LiDAR using performance analysis experiments based on measurement error and object detection experiments based on raster image.The experimental results show that the multi-lidar coupling system based on this paper has advantages and disadvantages in performance with HDL32 E LiDAR,but the object detection accuracy is higher and there’s practical application value.