Temporal Relaxation of STL Specifications in Distributed Multi-Agent Systems

We investigate a planning problem in a time-critical Multi-Agent System (MAS), where each agent is assigned with a local task formulated as a Signal Temporal Logic (STL) specification. Due to task coupling, the temporal constraints of some STL tasks should be relaxed to maintain the satisfaction of the edited tasks. An algorithm with a Distributed Model Predictive Control (DMPC) framework and an estimation mechanism of temporal relaxation are developed to ensure the satisfaction with minimized temporal relaxation. The efficacy and efficiency of the proposed algorithm are demonstrated via a simulation compared with an optimal baseline method using mixed integer programming encoding.