Deadlock Resolution of Connected Multi-Agent Systems using Hierarchical Control
Kunal Garg, Sera Hamilton, Chuchu Fan · 2024
Multi-agent robotic systems often require control design for a multi-objective problem, such as maintaining a safe distance from other agents as well as obstacles, maintaining network connectivity for building team knowledge, and completing team objectives for performance. Such problems are intractable in the centralized framework for large-scale systems. Thus, a distributed framework is necessary where each agent only requires its neighbors’ information while being able to contribute towards completing the team objective. However, a decentralized control framework often leads to a sub-optimal solution, resulting in the system getting stuck in local minima or a deadlock. This paper addresses the issue of deadlock resolution via a hierarchical control framework. We propose a high-level planner for temporary goal assignment and a lowlevel controller that drives the agents to their assigned goals. The proposed framework is distributed in nature, making it scalable to large-scale multi-agent systems. We perform extensive simulation and experimental case studies to demonstrate the efficacy and need for such a hierarchical control framework.