Routing for Bridge Inspecting Robots Using a Metaheuristic Genetic Algorithm

Deteriorating bridge infrastructure accounts for billions per year in inspection costs worldwide and recent advances in robotics and autonomy for bridge inspection can significantly decrease these costs. However, generating optimal tours for inspection robots to cover all members of a bridge truss maps to the well known but NP-hard Min-Max k-Chinese Postman arc-routing problem. We thus attack this problem with a new meta-heuristic genetic algorithm that quickly produces near-optimal balanced tours for k robots. Meta-heuristic genetic algorithm produced tour quality is statistically indistinguishable from best know results on common benchmarks. Scaling up to real-world bridge sizes, our genetic algorithm produces significantly better (15.24%) tours in a fraction of the time (0.05) compared to a prior genetic algorithm approach using a direct encoding. These results show the potential of our new approach for the broad class of arc-routing problems and specifically for quickly generating high-quality tours for robot-assisted real-world bridge inspection tasks.