We consider multi-robot applications, where a team of robots can ask for the intervention of a human operator to handle difficult situations. As the number of requests grows, team members will have to wait for the operator attention, hence the operator becomes a bottleneck for the system. Our aim in this context is to make the robots learn cooperative strategies to decrease the idle time of the system by modeling the operator as a shared resource. In particular, we consider a balking queuing model where robots decide whether or not to join the queue and use multi-robot learning to estimate the best cooperative policy. In more detail, we formalize the problem as Decentralized Markov Decision Process and provide a suitable state representation, so to apply an independent learners approach. We evaluate the proposed method in a robotic water monitoring simulation and empirically show that our approach can significantly improve the team performance, while being computationally tractable.

Additional Metadata
Keywords Autonomous surface vessels, Cooperative learning in MRS, Human-robot interaction, Multi-robot systems cooperation
Persistent URL dx.doi.org/10.1007/s10514-019-09877-w
Journal Autonomous Robots
Raeissi, M.M, & Farinelli, A. (2019). Cooperative queuing policies for effective scheduling of operator intervention. Autonomous Robots, 1–10. doi:10.1007/s10514-019-09877-w