Health-Aware Multi-UAV Planning using Decentralized Partially Observable Semi-Markov Decision Processes

Decentralized planning for large teams of robots is challenging, requiring control and coordination of the multi-robot team in settings with noisy sensors and stochastic transition dynamics. Although the Decentralized Partially Observable Markov Decision Process (Dec-POMDP) provides a general representation of this problem, solving it is often computationally infeasible for real-world domains. Recent work has focused on an extended representation of this framework, relying on usage of macro-actions (high-level, temporallyextended actions). This paper uses a macro-action based, Semi-Markovian decentralized framework for multi-UAV coordination in a health-aware setting. We also present hardware experiments for a constrained package delivery domain with asynchronous decision-making, demonstrating viable usage of this framework in real-world settings with continuous state spaces.