Communication-Aware Multi-Agent Metareasoning for Decentralized Task Allocation

Metareasoning refers to reasoning about one’s own decision making process. This paper considers metareasoning the process in multi-agent settings. We present a approach that enables team select which task allocation algorithm use as function of changing communication quality level. Given set algorithms, we synthesize policy prescribes best among predefined algorithms for given Since each agent runs same policy, (or part team) will collectively switch between observed level communication. apply reactive synthesis generate from high-level specifications written Linear Temporal Logic encoding agents’ switching behavior with respect state environment. perform experiments simulation identify performing under different levels. The environment is modeled using Rayleigh fading model and estimation done through exchange heartbeat messages agents. test our three types scenarios: search & rescue, fire monitoring, ship protection scenarios. For scenario, demonstrate achieved better performance either max distance traveled, number transmitted or both compared running any single algorithm.