Coverage path planning for legged robots in unknown environments

In applications like search and rescue, industrial inspection, or construction site mapping, robots may be used to explore and map unknown areas using onboard sensing. This paper presents a novel coverage path planning (CPP) algorithm for autonomous exploration with legged robots. The algorithm decomposes the region of interest into cells by detecting landmarks in the environment. Each cell is covered using a zig-zag motion pattern. The landmark detection is based on the traversability information of the surrounding area. The detection mechanism can be applied to obstacles with any shape, and it is robust against measurement uncertainty. Unlike existing CPP algorithms, the proposed algorithm takes into account the traversability of the surrounding terrain, and it is applicable to rough terrain environments where obstacle detection is a non-trivial problem. Completeness of exploration and the trade off between completeness and efficiency are discussed. Two simulations of the proposed algorithm were conducted using the open source robotics simulator Gazebo, with the legged robot ANYmal [1] as robotic platform. The results demonstrate the feasibility of the proposed algorithm for exploring unknown, rough terrain environments.