An energy aware Q-learning framework for comprehensive coverage path planning in unknown complex environments.

Area of Science:

• Robotics and Intelligent Systems
• Artificial Intelligence
• Search and Rescue Operations

Background:

• Conventional coverage path planning (CPP) algorithms face limitations in dynamic, unpredictable post-disaster environments.
• Robotic path planning requires adaptability for efficient search and rescue operations in uncertain terrains.

Purpose of the Study:

• To introduce PERM-QN (Q-learning with priority experience replay and memory network), an intelligent path planning algorithm.
• To enable energy-aware, complete area coverage for robots in complex, obstacle-laden post-disaster environments.

Main Methods:

• PERM-QN integrates a dynamic weight reward function for adaptive balancing of coverage, path length, and energy consumption.
• Priority experience replay mechanism accelerates learning convergence by prioritizing high-value past experiences.
• A memory network expedites route planning in similar terrain regions, reducing redundant exploration.

Main Results:

• PERM-QN demonstrated more efficient and comprehensive area exploration compared to traditional methods in simulated post-disaster environments.
• The algorithm maintained robust performance across varying environmental complexities.
• PERM-QN achieved effective energy-aware coverage in uncertain terrains.

Conclusions:

• PERM-QN offers an effective solution for robotic path planning in complex, dynamic post-disaster scenarios.
• The intelligent algorithm enhances the capabilities of robots in search and rescue missions.
• This approach improves exploration efficiency and completeness while managing energy consumption.