Workspace Modeling and Path Planning for Truss Structure Inspection by Unmanned Aircraft

Small unmanned aircraft systems can increase efficiency and reduce risk to humans during the inspection of truss-supported structures such as steel bridges. This paper describes a method to mathematically represent a truss and subsequently to plan an efficient collision-free inspection path. The algorithm checks user-defined perspectives for feasibility and redefines any infeasible perspectives. A deterministic roadmap is then generated as the complete graph over these perspectives, and some additional nodes are generated near the joints of the structure. A traveling salesman problem (TSP) is solved to find an efficient inspection path that tours all perspective points, and a local A* path planner then refines the inspection path to circumvent obstructions. The TSP is re-solved, with an updated cost based on local detours; and the process iterates until a feasible TSP solution is found. The "lazy" approach to global and local planning, for which paths are checked for feasibility after the fact and amended if necessary, ensures quick convergence to an efficient inspection path. Simulation results show the functionality of the algorithm. Comparison with a probabilistic roadmap method indicates the proposed algorithm's efficiency.