The paper addresses the problem of multi-agent navigation, where a group of agents must reach a set of target positions, but it is not necessary for a specific agent to reach a particular target. Unlike existing centralized methods, this work proposes a fully decentralized approach in which agents independently select targets and actions based on local observations and communication. To ensure consistent target assignment, new algorithms have been developed using a target and priority swapping procedure. This procedure has been applied to both discrete and continuous space representations. It has been proven that the proposed approaches guarantee the finding of a solution under certain conditions. Experimental results demonstrate that the proposed methods effectively handle various highly complex tasks.
Math-Net.Ru: https://www.mathnet.ru/eng/iipr651
HSE University: https://publications.hse.ru/en/articles/1134107257
Dergachev, Stepan. Decentralized navigation of multiple interchangeable agents // Artificial Intelligence and Decision Making, 2025, Issue 4, pp. 76–92.