A hybrid motion planning framework for three-wheeled mobile robots based on improved A* and nonlinear model predictive control

Abstract

This paper proposes a hybrid motion planning framework for three-wheeled mobile robots (3WMRs), integrating an improved A-star (A*) algorithm for global planning with Nonlinear Model Predictive Control (NMPC) for local planning. The improved A* algorithm is designed with a two-stage adaptive heuristic strategy that combines Manhattan and Octile distances. This approach significantly reduces the number of expanded nodes during the search process while mitigating the tendency of trajectories to graze obstacle boundaries through a safety buffer expansion mechanism within the grid map construction. The resulting global trajectory is smoothed using Cubic Spline interpolation and serves as the reference trajectory for the NMPC controller at the local planning layer. This ensures precise trajectory tracking, smooth motion, and strict adherence to kinematic constraints, non-holonomic constraints, workspace limits, and the robot's control limits. Furthermore, the paper proposes a real-time replanning mechanism based on the improved A* algorithm to handle sudden obstacles in dynamic environments. The effectiveness of the proposed method is validated through simulation scenarios in MATLAB R2023b across both static and dynamic environments, including quantitative comparisons with classical algorithms such as Dijkstra and conventional A*. The results demonstrate that the proposed method substantially decreases the number of expanded nodes while ensuring reliable collision avoidance and high tracking accuracy. These findings confirm the efficiency, robustness, and practical feasibility of the proposed hybrid motion planning framework for mobile robots in complex environments.