On optimization of task assignment in multi-agent systems (Proposal of practical solution search method with reduced computational load and application to mid-air retrieval of low-speed descending object)

This study discusses the concept of resilient multi-agent systems and algorithms for their realization. Specifically, we focus on improving the optimization algorithm for task assignments and applying it to a mid-air retrieval mission. In particular, this study highlights the mid-air retrieval of a low-speed descending object, supported by a parachute, by one of three unmanned aerial vehicles (UAVs) that constitute a multi-agent system as a representative case. The results demonstrate that the optimal task assignments can be achieved within practical computation times. This capability enables prompt and appropriate adjustments to task assignments in response to dynamic changes in situations and environments, marking a significant step toward realizing practical resilient multi-agent systems. Different from the conventional methods such that they pursue mathematically rigorous optimal solutions, this study aims to obtain solutions suitable for practical applications within shorter time frames. Therefore, we propose a method to derive optimal task assignments based on approximate trajectory planning, as verified through numerical simulations. These simulations demonstrate that, in the case study, one of the UAVs can successfully capture a low-speed descending object with feasible maneuvers. In this paper, the issue of deriving exact optimal trajectories using the obtained approximate trajectories as initial solutions still remains. This issue is one of the important subjects in this study and we will discuss it in another paper.