Adaptive Coverage Path Planning Policy for a Cleaning Robot with Deep Reinforcement Learning

This paper presents an adaptive policy for coverage path planning for a cleaning robot in 2D environments based on reinforcement learning. We applied an actor-critic model and a simulator to make a robot learn a path planning policy. In the view of consumer electronics, our objective function is designed to generate the minimum energy path. We used a real cleaning robot called R9 made by LG to evaluate our algorithm. Compared with a rule-based algorithm and other learning-based algorithms, our algorithm is probably more efficient in the view of energy saving.