Solving the Inspection Problem via Colored Traveling Salesman Tours

This paper describes an effective approach to solve the inspection problem via colored traveling salesman tours. Given an environment with obstacles and regions to be inspected, the approach plans a collision-free and dynamicallyfeasible motion trajectory that enables the vehicle to inspect all the regions of interest while reducing the distance traveled. A key aspect of the approach is the transformation of the inspection task into a colored traveling salesman problem (CTSP). This is achieved by generating a number of inspection points on the medial axis of each region in order to increase visibility and by grouping the inspection points according to their color. We also contribute a fast CTSP solver based on Monte-Carlo search to effectively find low-cost colored tours. Finally, a vehicle controller is employed in order to follow the planned tour. Experiments using different environments and inspection tasks provide promising validation.