Coverage Path Planning Using Actor-Critic Deep Reinforcement Learning

Area of Science:

• Robotics
• Artificial Intelligence
• Machine Learning

Background:

• Mobile robot exploration is crucial but faces challenges in complete environmental coverage.
• Existing coverage path planning methods remain an open problem despite advancements.

Purpose of the Study:

• To propose a deep reinforcement learning framework for mobile robot coverage path planning.
• To train and evaluate actor-critic algorithms for efficient environment exploration.

Main Methods:

• Utilized deep reinforcement learning with Advantage Actor-Critic (A2C) and Proximal Policy Optimization (PPO) algorithms.
• Defined environment states, observations, and reward functions tailored for robot exploration.
• Trained policies for a mobile robot navigating an environment with obstacles.

Main Results:

• Optimized policies were generated using both A2C and PPO algorithms.
• Evaluation demonstrated the effectiveness of actor-critic methods in guiding robot exploration.
• The proposed framework enables efficient coverage of unknown terrains.

Conclusions:

• Actor-critic reinforcement learning methods are capable of producing effective policies for mobile robot coverage path planning.
• This approach offers a viable solution for robots to explore and map new environments autonomously.