A substation robot path planning algorithm based on deep reinforcement learning enhanced by ant colony optimization.

Area of Science:

• Robotics
• Artificial Intelligence
• Optimization Algorithms

Background:

• Substation robots require advanced path planning due to complex electromagnetic fields, dense equipment, and safety demands.
• Existing methods struggle with the unique challenges of substation environments for inspection and maintenance.

Purpose of the Study:

• To develop an improved path planning algorithm for substation robots.
• To enhance operational efficiency and safety in substation inspection and maintenance tasks.

Main Methods:

• A synergistic framework combining deep reinforcement learning (DRL) with ant colony optimization (ACO).
• Pheromone-guided exploration strategy to reduce ineffective pathfinding.
• Sample screening mechanism using ACO path experience to boost Q-network training.
• Dynamic adjustment of decision weights for a gradual shift from heuristic to autonomous learning.

Main Results:

• Achieved 24% higher sample efficiency compared to baseline DRL algorithms.
• Reduced average path length by 18% and demonstrated superior dynamic obstacle avoidance.
• Field validation showed a 14.8% improvement in task completion rate in a real substation.

Conclusions:

• The proposed DRL-ACO algorithm significantly outperforms state-of-the-art methods in substation path planning.
• The hybrid approach enhances sample efficiency, path optimality, and obstacle avoidance capabilities.
• Demonstrated practical effectiveness and improved task completion in real-world substation environments.