Towards Human-Like RL: Taming Non-Naturalistic Behavior in Deep RL via Adaptive Behavioral Costs in 3D Games

TL;DR

This paper introduces ABC-RL, a reinforcement learning method that encourages human-like behavior in agents by dynamically penalizing unnatural actions, achieving natural gameplay without sacrificing performance.

Contribution

The paper presents a novel adaptive behavioral cost framework using augmented Lagrangian to produce human-like behavior in deep RL agents while maintaining competitive performance.

Findings

Reduces unnatural behaviors like shaking and spinning in 3D game agents.

Maintains the same performance level as unconstrained RL agents.

Demonstrates effectiveness across DMLab-30 and Unity ML-Agents environments.

Abstract

In this paper, we propose a new approach called Adaptive Behavioral Costs in Reinforcement Learning (ABC-RL) for training a human-like agent with competitive strength. While deep reinforcement learning agents have recently achieved superhuman performance in various video games, some of these unconstrained agents may exhibit actions, such as shaking and spinning, that are not typically observed in human behavior, resulting in peculiar gameplay experiences. To behave like humans and retain similar performance, ABC-RL augments behavioral limitations as cost signals in reinforcement learning with dynamically adjusted weights. Unlike traditional constrained policy optimization, we propose a new formulation that minimizes the behavioral costs subject to a constraint of the value function. By leveraging the augmented Lagrangian, our approach is an approximation of the Lagrangian adjustment, which handles the trade-off between the performance and the human-like behavior. Through experiments conducted on 3D games in DMLab-30 and Unity ML-Agents Toolkit, we demonstrate that ABC-RL achieves the same performance level while significantly reducing instances of shaking and spinning. These findings underscore the effectiveness of our proposed approach in promoting more natural and human-like behavior during gameplay.