Multi-robot obstacle avoidance in dynamic environments using opinion-driven CBFs

We present a cooperative decentralized obstacle avoidance approach for a group of ground-based robots moving in an environment with dynamic obstacles. The objectives of this work are to ensure safety, on the one hand, and to embed coordinated behavior, on the other. While Control Barrier Functions (CBFs) have proven to be effective while guaranteeing safety, they often lead to overly conservative behavior and deadlocks in dense environments. In this paper, we propose a framework in which agents use the dynamics of opinion for decision making. Since CBFs do not consider any cooperative strategies for avoiding collisions and deadlocks in the presence of dynamic obstacles, we introduce two heuristics, which effect the opinion evolution model: the local attention mechanism, focusing on reacting to collisions with nearby obstacles; the global attention mechanism, which further facilitates cooperation between agents. The simulation results demonstrate that our proposed method reduces deadlocks, improves trajectory smoothness, and shortens the task completion time compared to the only CBF strategies.