Global-Initialization-Based Model Predictive Control for Mobile Robots Navigating Nonconvex Obstacle Environments

Abstract

This paper proposes a nonlinear model predictive control (MPC) framework initialized using an initial-guess particle swarm optimization (IG-PSO) algorithm for mobile robots navigating in environments with nonconvex obstacles. The proposed method is designed to address the local minimum problem inherent in conventional optimization-based MPC by incorporating a PSO-based global search method to generate effective initial guesses. In addition, a grid-based representation of the nonconvex obstacles is implemented to systematically define the collision avoidance constraints within the MPC formulation. The proposed method was validated in real-time simulations using the Robot Operating System (ROS) and the Gazebo physics simulator. The results demonstrate that the proposed MPC initialized by IG-PSO generates collision-free trajectories that avoid local minima and track the desired reference trajectory in environments with nonconvex obstacles. Compared with conventional IPOPT-based MPC, the proposed method exhibited improved performance in the tested scenario. The proposed method also maintains real-time control capabilities by selectively activating the IG-PSO algorithm only as required. The findings of this study demonstrate the potential of the proposed framework for robust and efficient trajectory planning in complex, nonconvex obstacle environments.