Model Predictive Control of UAVs with Actor-Critic and Artificial Potential Field

UAV technology is developing rapidly and is widely used in various fields. However, achieving efficient control in complex dynamic environments remains a major challenge. To this end, this paper proposes a UAV model predictive control (MPC) method that combines an actor-critic algorithm with an artificial potential field. The method enhances the UAV's environment sensing and path planning capabilities by combining the artificial potential field. By combining the short-term prediction optimisation of the MPC with the exploratory and end-to-end training capabilities of the Actor-Critic algorithm, the proposed method efficiently manages short-term strategies through the Actor network-based MPC and long-term predictions against the Critic network. This approach unifies the advantages of artificial potential fields, model-based MPC and model-free Actor-Critic algorithms. Experimental results show that the proposed method performs well in path planning and forecast tracking in dynamic multi-obstacle environments.