Unmanned aerial vehicle adaptive route planning and dynamic obstacle avoidance method and system for power distribution line inspection

Abstract

The application provides a kind of unmanned aerial vehicle adaptive route planning and dynamic obstacle avoidance method and system for distribution line inspection.The method first constructs a special visual data set and optimizes the target detection network,improves the detection accuracy by fusing double attention mechanism and trainable weight feature fusion module;Deploy a multi-sensor fusion system to realize three-dimensional positioning of the target,combined with video stream filtering to suppress detection jitter;Design a deep reinforcement learning obstacle avoidance and navigation model based on the TD3 framework,combine GRU and attention mechanism to process sequence state,achieve safe and efficient navigation through multi-object reward function;Establish a priority decision mechanism to execute feature recognition,pose adjustment and dynamic obstacle avoidance simultaneously,use model predictive control to realize closed-loop path planning.The system shows stable adaptive inspection and obstacle avoidance capability in complex line environment.

Images