To address the problem that wheeled unmanned vehicles cannot complete the expected navigation task due to the non-compliance with vehicle kinematics when using traditional autonomous navigation systems, this paper modularizes the design of autonomous navigation system based on the ROS platform, adopts graph optimization-based cartographer algorithm to realize the functions of map building and localization, and uses ARA*-based global path planning SBPL_Lattic_Planner and elastic band-based local path planning TEB_Local_Planner to plan the path and control the vehicle to complete the autonomous navigation function. Lattic_Planner based on ARA* algorithm and TEB_Local_Planner based on elastic band to plan the path and control the vehicle to fulfill the autonomous navigation function. In addition, a test experiment was set up for the kinematic characteristics of the vehicle and compared with the conventional autonomous navigation system. The results show that the new autonomous navigation system is able to plan a path that meets the kinematic characteristics of the vehicle and effectively copes with sudden obstacles and narrow spaces, which meets the requirements of intelligent wheeled unmanned vehicles for autonomous navigation.