基于ROS与YOLOv5s的智能车障碍物检测导航系统的设计

Design of Intelligent Vehicle Obstacle Detection and Navigation System based on ROS and YOLOv5s

针对智能车障碍物检测与自主导航任务中存在的识别准确率差、检出率低,以及自主导航路径规划器稳定性差等问题,设计一种基于ROS实验平台的障碍物检测与自主导航路径规划系统。系统以YOLOv5s作为障碍物检测算法框架,以A^(*)算法与TEB算法融合作为自主导航算法框架,改善了智能车障碍物检测精度低与路径规划不稳定的问题。实验结果表明,搭载该系统的智能车能够完成障碍物检测、自主导航的任务,路径规划平均成功率达到96.67%,障碍物检测准确率在92%以上,综合任务成功率在90%以上,具有障碍物检测准确率高,自主导航路径规划稳定性强的特性。

Aiming at the problems of poor recognition accuracy and low detection rate in obstacle detection and autonomous navigation tasks of intelligent vehicles,as well as poor stability of autonomous navigation path planner,this paper designed an obstacle detection and autonomous navigation path planning system based on ROS experimental platform.This system uses YOLOv5s as the obstacle detection algorithm framework and the fusion of A*algorithm and TEB algorithm as the autonomous navigation algorithm framework,which improves the accuracy of obstacle detection and makes path planning of intelligent vehicles more stable.The results show that the intelligent vehicle equipped with the system can complete the tasks of obstacle detection and autonomous navigation,and the average success rate of path planning is 96.67%,the accuracy rate of obstacle detection is more than 92%,and the success rate of comprehensive tasks is more than 90%.It has the characteristics of high accuracy rate of obstacle detection and strong stability of autonomous navigation path planning.