Coverage Path Planning Using Reinforcement Learning-Based TSP for hTetran-A Polyabolo-Inspired Self-Reconfigurable...

Area of Science:

• Robotics
• Artificial Intelligence
• Computational Geometry

Background:

• Deploying cleaning robots faces challenges in achieving complete area coverage due to fixed robot forms.
• Reconfigurable robot systems offer a solution for optimal area coverage by adapting shapes to environmental needs.
• Efficient navigation strategies are crucial for extending coverage range and minimizing energy consumption.

Purpose of the Study:

• To present a complete path planning (CPP) strategy for the hTetran reconfigurable tiling robot.
• To optimize robot shape configuration and navigation trajectories simultaneously using reinforcement learning.
• To enhance area coverage efficiency in terms of energy and time spent.

Main Methods:

• Developed a complete path planning (CPP) approach for the hTetran robot based on a Traveling Salesperson Problem (TSP)-inspired reinforcement learning (RL) optimization.
• Employed the Proximal Policy Optimization (PPO) algorithm to train the RL model for maximizing coverage rewards.
• Compared the proposed RL-TSP-based CPP with conventional TSP solutions using Ant Colony Optimization (ACO) for tiled robots.

Main Results:

• The RL-TSP-based CPP for hTetran successfully generated robot shapes and sequential trajectories.
• The proposed method achieved optimal Pareto trajectories, enhancing navigation efficiency.
• Comparative analysis showed reduced energy and time consumption compared to conventional ACO-based TSP methods.

Conclusions:

• The developed RL-TSP-based CPP effectively addresses the area coverage challenge for reconfigurable robots like hTetran.
• This approach optimizes both robot configuration and path planning for superior energy and time efficiency.
• The findings suggest a promising direction for intelligent robotic systems in complex environment coverage.