Action Guidance: Getting the Best of Sparse Rewards and Shaped Rewards for Real-time Strategy Games

TL;DR

This paper introduces action guidance, a novel reinforcement learning technique that improves training efficiency in sparse reward RTS games by ensuring agents ultimately optimize the true objective, overcoming reward shaping drawbacks.

Contribution

The paper proposes action guidance, a new method that combines reward shaping benefits with true objective optimization in sparse reward environments.

Findings

Effective in μRTS game simulator

Maintains sample efficiency of reward shaping

Ensures agents optimize true objectives

Abstract

Training agents using Reinforcement Learning in games with sparse rewards is a challenging problem, since large amounts of exploration are required to retrieve even the first reward. To tackle this problem, a common approach is to use reward shaping to help exploration. However, an important drawback of reward shaping is that agents sometimes learn to optimize the shaped reward instead of the true objective. In this paper, we present a novel technique that we call action guidance that successfully trains agents to eventually optimize the true objective in games with sparse rewards while maintaining most of the sample efficiency that comes with reward shaping. We evaluate our approach in a simplified real-time strategy (RTS) game simulator called $\mu$RTS.