Adaptive Operation Path Planning Method of High-Voltage Line Fire-Fighting Robot Based on Reinforcement Learning

The high-voltage line fire-fighting robot has problems such as poor real-time path planning, insufficient obstacle avoidance accuracy, and low operation efficiency in complex live working scenes. This study constructs a high-precision three-dimensional environment model and establishes a safe distance constraint. A hierarchical MDP framework is adopted: the state space integrates joint angles, terminal posture, 128-dimensional obstacle point cloud and electric field strength; the action space is the joint angular velocity. Improved DDPG algorithm: Actor network is embedded in GCN to process obstacle point cloud, Critic uses TD3 double Q network, and PER mechanism optimizes sample collection. This method achieves an average planning time of 1.285 seconds, $97.44 \%$ obstacle avoidance success rate and 3.79 mm trajectory tracking error in a ± 500kV simulation environment, effectively ensuring the safety and efficiency of live-line operations.