Fighting Copycat Agents in Behavioral Cloning from Observation Histories

TL;DR

This paper introduces an adversarial method to improve imitation learning by reducing the copycat problem, where agents predict previous actions instead of the next, especially in partially observed settings.

Contribution

It proposes a novel adversarial feature learning approach to mitigate causal confusion in behavioral cloning from observation histories.

Findings

Significant performance improvements across various tasks

Effective reduction of copycat behavior in imitation learning

Enhanced robustness in partially observed environments

Abstract

Imitation learning trains policies to map from input observations to the actions that an expert would choose. In this setting, distribution shift frequently exacerbates the effect of misattributing expert actions to nuisance correlates among the observed variables. We observe that a common instance of this causal confusion occurs in partially observed settings when expert actions are strongly correlated over time: the imitator learns to cheat by predicting the expert's previous action, rather than the next action. To combat this "copycat problem", we propose an adversarial approach to learn a feature representation that removes excess information about the previous expert action nuisance correlate, while retaining the information necessary to predict the next action. In our experiments, our approach improves performance significantly across a variety of partially observed imitation learning tasks.