Scalable, Decentralized Multi-Agent Reinforcement Learning Methods Inspired by Stigmergy and Ant Colonies

TL;DR

This paper introduces a scalable, decentralized multi-agent reinforcement learning method inspired by ant colonies, aiming to improve coordination and scalability in complex environments, demonstrated through a path planning task involving environment modification.

Contribution

The work presents a novel ant-colony-inspired decentralized algorithm combining RL and stigmergy, addressing scalability and non-stationarity in multi-agent systems.

Findings

Algorithm is scalable to many agents

Promising success in a specific environment

Limited generalization and performance due to simplicity

Abstract

Bolstering multi-agent learning algorithms to tackle complex coordination and control tasks has been a long-standing challenge of on-going research. Numerous methods have been proposed to help reduce the effects of non-stationarity and unscalability. In this work, we investigate a novel approach to decentralized multi-agent learning and planning that attempts to address these two challenges. In particular, this method is inspired by the cohesion, coordination, and behavior of ant colonies. As a result, these algorithms are designed to be naturally scalable to systems with numerous agents. While no optimality is guaranteed, the method is intended to work well in practice and scale better in efficacy with the number of agents present than others. The approach combines single-agent RL and an ant-colony-inspired decentralized, stigmergic algorithm for multi-agent path planning and environment modification. Specifically, we apply this algorithm in a setting where agents must navigate to a goal location, learning to push rectangular boxes into holes to yield new traversable pathways. It is shown that while the approach yields promising success in this particular environment, it may not be as easily generalized to others. The algorithm designed is notably scalable to numerous agents but is limited in its performance due to its relatively simplistic, rule-based approach. Furthermore, the composability of RL-trained policies is called into question, where, while policies are successful in their training environments, applying trained policies to a larger-scale, multi-agent framework results in unpredictable behavior.