Publikation

Verifying Safety Properties of Robotic Plans operating in Real-World Environments via Logic-based Environment Modeling

Tim Meywerk, Marcel Walter, Vladimir Herdt, Jan Kleinekathöfer, Daniel Große, Rolf Drechsler

In: 9th International Symposium On Leveraging Applications of Formal Methods, Verification and Validation (ISoLA). International Symposium On Leveraging Applications of Formal Methods, Verification and Validation (ISoLA-2020) October 20-24 Rhodes Greece 2020.

Abstrakt

These days, robotic agents are finding their way into thepersonal environment of many people. With robotic vacuum cleanerscommercially available already, comprehensive cognition-enabled agentsassisting around the house autonomously are a highly relevant researchtopic. To execute these kinds of tasks in constantly changing environ-ments, complex goal-driven control programs, so-called plans, are required. They incorporate perception, manipulation, and navigation capabilities among others. As with all technological innovation, consequently,safety and correctness concerns arise.In this paper, we present a methodology for the verification of safetyproperties of robotic plans in household environments by a combina-tion of environment reasoning usingDiscrete Event Calculus(DEC) andSymbolic Executionfor effectively handling symbolic input variables (e. g.object positions). We demonstrate the applicability of our approach inan experimental evaluation by verifying safety properties of robotic planscontrolling a two-armed, human-sized household robot packing and un-packing a shelf. Our experiments demonstrate our approach’s capabilityto verify several robotic plans in a realistic, logically formalized environment.

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence