Randomized planning for short inspection paths

Abstract

Addresses the following inspection problem: given a known workspace and a robot with vision capabilities compute a short path path for the robot such that each point on boundary of the workspace is visible from some point on the path. Autonomous inspection, such as by a flying camera, or a virtual reality architectural walkthrough, could be guided by a solution to the above inspection problem. Visibility constraints on both maximum viewing distance and maximum angle of incidence are considered to better model real sensors. An algorithm is presented for planar workspaces which operates in two steps: selecting art gallery-style guards and connecting them to form an inspection path. Experimental results for this algorithm are discussed. Next, the algorithm is extended to three dimensions and inspection paths are shown.