MPC Trajectory Tracking and Obstacle Avoidance Control Based on Real-Time Polygon Modeling

This paper proposes a trajectory tracking control framework based on model predictive control (MPC), which enables the controlled object to perform obstacle avoidance while considering its dynamic model. By scanning the environment around the mobile robot, depth information point clouds are obtained and then modeled as convex hulls. The Gilbert-Johnson-Keerthi (GJK) algorithm is used to simplify the distance problem between the obstacle convex hull and the mobile robot convex hull. The resulting constraint equations are added to the optimization problem of MPC-based trajectory tracking control. This approach compensated an obstacle avoidance safeguard at the final level of motion control, thereby reducing the planning complexity for higher-level tasks such as path planning. Finally, an unmanned surface vessel (USV) control problem was used to conduct simulation experiments, verifying the real-time obstacle avoidance capability. The practical prospects of this framework were explored in terms of feasibility and computational efficiency.