Designing an obstacle detection and avoidance system using a 2D LIDAR and a stereo camera for autonomous vehicles

In this study, some software has been developed by applying processing, and deep learning methods to detect obstacles, to map the environment of the autonomous car and to plan of car's path with the Nvidia Jetson TX2, Lidar, and ZED Camera. The operating system of the tool is Linux-based operating system (Ubuntu). CUDA, VisionWorks, and OpenCV software will be installed on this operating system by installing NVIDIA's JetPack software package. Drivers will be installed for LIDAR and camera to work. Robot Operating System (ROS) for simulation and control of mobile robots and ZED SDK for ZED stereo camera software support will be uploaded. In this study, Hector Slam Mapping and 3 different Path Planning Algorithms were used for autonomous vehicle design with Lidar and stereo camera. By applying mapping in Rviz, it was observed that map information was taken from Lidar instantly, and the next location could be determined by using route estimation. Finally, testing and simulation was performed between the start and end points with the Gazebo simulation tool, and it was observed that 3 different algorithms performed the road planning on a different route and the results were compared. Keywords: Object Detection, Autonomous Vehicle, Path Planning, Motion Control, Deep Learning