Robust Robotic Assembly through Contingencies, Plan Repair and Re-Planning

Abstract

Enabling mobile robots to assemble large structures in constrained environments requires planning systems that are both capable of dealing with high complexity and can provide robust execution in the face of run-time failures. We achieve execution robustness through exception handling capabilities that are seamlessly integrated throughout the planning system. Having these recovery mechanisms in place allows us to leverage their capabilities to compensate for problems introduced by approximations made during planning. Turning an apparent problem into an opportunity, we are able to plan complex assembly tasks and execute them robustly without the computational cost associated with more sophisticated planners and apply some of the savings toward recovering from unforeseen run-time errors. We show results where simple planning strategies paired with exception-handling are able to achieve the same outcomes (and in less time) as more elaborate methods would.