Synthesis of Provably Correct Autonomy Protocols for Shared Control

We synthesize shared control protocols subject to probabilistic temporal logic specifications. Specifically, we develop a framework in which human and an autonomy protocol can issue commands carry out certain task. blend these into joint input robot. model the interaction between robot as Markov decision process representing scenario. Using inverse reinforcement learning, obtain abstraction of human's behavior. use randomized strategies account for randomness decisions, caused by factors such complexity task specifications or imperfect interfaces. design ensure that resulting behavior satisfies given safety performance logic. Additionally, generate similar induced possible. solve underlying problem efficiently using quasiconvex programming. Case studies involving autonomous wheelchair navigation unmanned aerial vehicle mission planning showcase applicability our approach.