Oracles & Followers: Stackelberg Equilibria in Deep Multi-Agent Reinforcement Learning
Keywords: Multi-Agent Reinforcement Learning, Stackelberg Equilibrium, Security Games, Stackelberg Security Games
TL;DR: We present a novel theoretical framework for learning Stackelberg equilibria in multi-agent RL settings such as security games, and a new Meta-RL approach that greatly speeds up learning and scales to novel domains.
Abstract: Stackelberg equilibria arise naturally in a range of popular learning problems, such as in security games or indirect mechanism design, and have received increasing attention in the reinforcement learning literature. We present a general framework for implementing Stackelberg equilibria search as a multi-agent RL problem, allowing a wide range of algorithmic design choices. We discuss how previous approaches can be seen as specific instantiations of this framework. As a key insight, we note that the design space allows for approaches not previously seen in the literature, for instance by leveraging multitask and meta-RL techniques for follower convergence. We propose one such approach using contextual policies, and evaluate it experimentally on both standard and novel benchmark domains, showing greatly improved sample efficiency compared to previous approaches. Finally, we explore the effect of adopting algorithm designs outside the borders of our framework.
Submission Number: 15
Loading