Abstract

Multi-agent navigation in restricted environments presents significant challenges, since agents need to move to their goals in an efficient manner while avoiding collisions with both static and dynamic obstacles. Previously, C-Nav was proposed as a method that can effectively coordinate groups of agents in very restricted environments. In this work, we propose and evaluate three alternative communications models for C-Nav, that intend to provide more flexibility to the original method. The results of our experiments show that each of our proposed methods can lead to significant improvements over C-Nav in specific types of restricted environments.