Temporal Induced Self-Play for Stochastic Bayesian Games

TL;DR

This paper introduces TISP, a reinforcement learning framework that ensures effective strategies from any decision point in stochastic Bayesian games, improving scalability and performance over existing methods.

Contribution

The paper proposes TISP and TISP-PG, novel algorithms that find approximate Perfect Bayesian Equilibria in stochastic Bayesian games with better scalability and performance.

Findings

TISP-PG outperforms existing learning-based methods.

TISP-based algorithms can find approximate Perfect Bayesian Equilibria.

TISP is more scalable than traditional mathematical programming methods.

Abstract

One practical requirement in solving dynamic games is to ensure that the players play well from any decision point onward. To satisfy this requirement, existing efforts focus on equilibrium refinement, but the scalability and applicability of existing techniques are limited. In this paper, we propose Temporal-Induced Self-Play (TISP), a novel reinforcement learning-based framework to find strategies with decent performances from any decision point onward. TISP uses belief-space representation, backward induction, policy learning, and non-parametric approximation. Building upon TISP, we design a policy-gradient-based algorithm TISP-PG. We prove that TISP-based algorithms can find approximate Perfect Bayesian Equilibrium in zero-sum one-sided stochastic Bayesian games with finite horizon. We test TISP-based algorithms in various games, including finitely repeated security games and a grid-world game. The results show that TISP-PG is more scalable than existing mathematical programming-based methods and significantly outperforms other learning-based methods.