[2605.04939] Modular Reinforcement Learning For Cooperative Swarms

Computer Science > Robotics arXiv:2605.04939 (cs) [Submitted on 6 May 2026] Title:Modular Reinforcement Learning For Cooperative Swarms Authors:Erel Shtossel, Gal A. Kaminka View a PDF of the paper titled Modular Reinforcement Learning For Cooperative Swarms, by Erel Shtossel and 1 other authors View PDF HTML (experimental) Abstract:A cooperative robot swarm is a collective of computationally-limited robots that share a common goal. Each robot can only interact with a small subset of its peers, without knowing how this affects the collective utility. Recent advances in distributed multi-agent reinforcement learning have demonstrated that it is possible for robots to learn how to interact effectively with others, in a manner that is aligned with the common goal, despite each robot learning independently of others. However, this requires each robot to represent a potentially combinatorial number of interaction states, challenging the memory capabilities of the robots. This paper proposes an alternative approach for representing spatial interaction states for multi-robot reinforcement learning in swarms. A modular (decomposed) representation is used, where each feature of the state is handled by a separate learning procedure, and the results aggregated. We demonstrate the efficacy of the approach in numerous experiments with simulated robot swarms carrying out foraging. Subjects: Robotics (cs.RO); Artificial Intelligence (cs.AI) Cite as: arXiv:2605.04939 [cs.RO]   (or arXiv:2...