THE INTEGRATION OF ARTIFICIAL NEURAL NETWORKS FOR MOTION PLANNING IN THE LEARNING AND CONTROL OF MOTOR SKILLS

This study investigates motion planning methods by integrating artificial neural networks (ANNs) into the learning and control processes of motor skills, aiming to enhance robots’ autonomous navigation and path-planning capabilities in complex environments. Based on reinforcement learning, a motion planning algorithm is developed that leverages ANNs to construct the state space, action space, and reward functions, while employing neural networks to approximate the state-action value function. Through iterative updates of network parameters, the algorithm optimizes autonomous navigation and path planning for robots. Performance evaluations in a simulated environment show that, under the control of the proposed improved path planning algorithm, the robot’s maximum and minimum errors in reaching the target point are 0.56[Formula: see text]cm and 0.21[Formula: see text]cm, respectively. In comparison, the ANN-based and A*-based algorithms exhibit larger path errors, with maximum errors of 1.56[Formula: see text]cm and 2.03[Formula: see text]cm, respectively. Furthermore, the proposed method achieves successful autonomous navigation and path planning on a physical robot platform, demonstrating high accuracy and stability in complex environments. The main contribution of this study lies in significantly enhancing robots’ autonomous navigation and path planning performance, providing strong technical support for the intelligent development of robotic systems.