A topological coverage algorithm for mobile robots

In applications such as vacuum cleaning, painting, demining and foraging, a mobile robot must cover an unknown surface. The efficiency and completeness of coverage is improved via the construction of a map of covered regions while the robot covers the surface. Existing methods generally use grid maps, which are susceptible to odometry error and may require considerable memory and computation. This paper proposes a topological map and presents a coverage algorithm in which natural landmarks are added as nodes in a partial map. The completeness of the algorithm is argued. Simulation tests show over 99% of the surface is covered; 85% for real (Khepera) robot tests. The path length is about 10% worse than optimal in simulation tests, and about 20% worse than optimal for the real robot, which are within theoretical upper bounds for approximate solutions to travelling salesman based coverage problems. The proposed algorithm generates shorter paths and covers a wider variety of environments than topological coverage based on Morse decompositions.