Coverage Path Planning Using Left Hand Rule and Slave Reinforcement Learning

Coverage path planning (CPP) is a topic in robotics and automation with the goal for an agent to pass every point of travel in an area. Many interesting arguments have been offered to answer this problem, yet each answer has drawbacks, and researchers are challenged to decrease its flaws with their algorithms. This paper presents a combined method of the left-hand rule (LHR) method and the slave reinforcement learning (SRL) to answer the coverage path planning problem. By using LHR, we can divide the whole area into room partitions. When an area is passed by LHR, there may be a loop closing for the area to be partitioned as a room. After that, the agent will use the SRL to travel inside the partitioned area. Using LHR and SRL separately reduces the computing cost for the system to train SRL. Experimental results show that the proposed method achieves the coverage ratio to an average of 1.11 and the path efficiency to an average of 0.92 which shows effectivity in the usage of reinforcement learning for coverage path planning. The work is useful for smart navigation, smart mapping, and smart robotics.