Reinforcement Learning From Imperfect Corrective Actions And Proxy Rewards

TL;DR

This paper introduces ICoPro, a novel deep reinforcement learning algorithm that effectively combines imperfect proxy rewards with human corrective actions to improve policy alignment and sample efficiency.

Contribution

The paper proposes ICoPro, a new value-based RL method that integrates human corrective feedback with proxy rewards, including pseudo-labels, to enhance learning in imperfect settings.

Findings

ICoPro outperforms baseline methods in aligning with human preferences.

The method is more sample-efficient across various tasks.

It effectively handles different types of feedback imperfection.

Abstract

In practice, reinforcement learning (RL) agents are often trained with a possibly imperfect proxy reward function, which may lead to a human-agent alignment issue (i.e., the learned policy either converges to non-optimal performance with low cumulative rewards, or achieves high cumulative rewards but in undesired manner). To tackle this issue, we consider a framework where a human labeler can provide additional feedback in the form of corrective actions, which expresses the labeler's action preferences although this feedback may possibly be imperfect as well. In this setting, to obtain a better-aligned policy guided by both learning signals, we propose a novel value-based deep RL algorithm called Iterative learning from Corrective actions and Proxy rewards (ICoPro), which cycles through three phases: (1) Solicit sparse corrective actions from a human labeler on the agent's demonstrated trajectories; (2) Incorporate these corrective actions into the Q-function using a margin loss to enforce adherence to labeler's preferences; (3) Train the agent with standard RL losses regularized with a margin loss to learn from proxy rewards and propagate the Q-values learned from human feedback. Moreover, another novel design in our approach is to integrate pseudo-labels from the target Q-network to reduce human labor and further stabilize training. We experimentally validate our proposition on a variety of tasks (Atari games and autonomous driving on highway). On the one hand, using proxy rewards with different levels of imperfection, our method can better align with human preferences and is more sample-efficient than baseline methods. On the other hand, facing corrective actions with different types of imperfection, our method can overcome the non-optimality of this feedback thanks to the guidance from proxy reward.