Using Dynamical Systems Theory for MARL Stability Analysis

mircomusolesi · x · 2026-08-17

This paper integrates Multi-Agent Reinforcement Learning (MARL) with dynamical systems theory from an individual perspective. It models MARL training as a coupled stochastic dynamical system to capture agent interactions and environmental dynamics. The research uses dynamical systems theory to analyze stability, sensitivity, and inherent stochasticity of agent behavior for safe deployment.

Related event: RLC 2026 Paper Applies Dynamical Systems Lens to Multi-Agent RL(2 posts)→

Original post →

More from Research

Research channel →