MPC-based Optimization Design for 3D Collision Avoidance of a Mobile Manipulator Based-on Obstacle Velocity Estimation

This paper presents an optimization design of 3D dynamic obstacle avoidance for a mobile manipulator based on model predictive control (MPC). The design enables a mobile manipulator to achieve optimized 3D collision avoidance motion with shorter avoidance path and faster avoidance time. A 3D LiDAR is installed onboard the robot to acquire environmental point cloud and estimate obstacle velocity. The MPC is designed to track an initial 3D path of the mobile manipulator and avoid any static and dynamic obstacles in real time. Experimental results show that the proposed method can simultaneously avoid static and dynamic obstacles in 3D space. Compared with baseline algorithms without velocity estimation, the proposed method reduces the avoidance path length by 8.27% and path execution time by 13.79%.