TriLayer-Nav: A Tri-Layer Navigation Framework Integrating A*, Dynamic Window Approach, and Model Predictive Control for Differential-Drive Robots

This paper introduces TriLayer-Nav, a modular tri-layer hybrid navigational system that integrates global planning, reactive local planning, and predictive control optimization to achieve smooth, energy-efficient and robust navigation of differential-drive ground robots. The architecture employs the A* algorithm for the computation of a collision-free global path, the dynamic window approach (DWA) for reactive obstacle avoidance, and model predictive control (MPC) for refining the DWA commands and at the same time satisfying non-holonomic and actuator constraints. The different layers function in cascade (A* at a low frequency, DWA in real time, and MPC continuously refining the commands), with final wheel-level tracking being performed by a PID controller. TriLayer-Nav has undergone intensive simulation and validation in a physics-based environment exploiting the MuJoCo platform, allowing detailed modelling of rigid-body dynamics, frictional interactions, and actuator feedback. In the simulation protocol, a complete hierarchical allocation of commands was highlighted throughout the system and at the same time maintained real-time computation throughput. The findings show that TriLayer-Nav produces smoother paths with less curvature discontinuities, reduced control oscillations, better heading accuracy, and lower energy use, and with a success rate of 96.6%, while path efficiency is better than that of implementations that solely rely on A* or DWA. Layer interaction provides better response to environmental changes and sensor noise. It is also computationally viable to run the framework in real time by using a nonlinear MPC solver, and its modular hierarchy allows it to be implemented in structured and semi-structured environments on a large scale. To sum up, the findings deliver strong evidence that combining global planning, local reactive planner and predictive optimization is a significant improvement in enhancing the reliability, stability and energy efficiency of autonomous navigation. TriLayer-Nav is a generalizable and computationally efficient navigation solution that can be applicable to a wide range of ground robotic systems used in dynamic and uncertain environments.