Residual Reward Models for Preference-based Reinforcement Learning

TL;DR

This paper introduces Residual Reward Models (RRM) for preference-based reinforcement learning, which combine prior knowledge with learned rewards to improve convergence speed and performance in complex environments, including real robots.

Contribution

The paper proposes a novel Residual Reward Model framework that effectively integrates prior reward knowledge with learned preferences, enhancing RL performance and convergence speed.

Findings

RRM significantly improves RL performance across multiple tasks.

The method accelerates policy learning in real robot experiments.

RRM is effective with various types of prior rewards, including IRL and proxy rewards.

Abstract

Preference-based Reinforcement Learning (PbRL) provides a way to learn high-performance policies in environments where the reward signal is hard to specify, avoiding heuristic and time-consuming reward design. However, PbRL can suffer from slow convergence speed since it requires training in a reward model. Prior work has proposed learning a reward model from demonstrations and fine-tuning it using preferences. However, when the model is a neural network, using different loss functions for pre-training and fine-tuning can pose challenges to reliable optimization. In this paper, we propose a method to effectively leverage prior knowledge with a Residual Reward Model (RRM). An RRM assumes that the true reward of the environment can be split into a sum of two parts: a prior reward and a learned reward. The prior reward is a term available before training, for example, a user's ``best guess'' reward function, or a reward function learned from inverse reinforcement learning (IRL), and the learned reward is trained with preferences. We introduce state-based and image-based versions of RRM and evaluate them on several tasks in the Meta-World environment suite. Experimental results show that our method substantially improves the performance of a common PbRL method. Our method achieves performance improvements for a variety of different types of prior rewards, including proxy rewards, a reward obtained from IRL, and even a negated version of the proxy reward. We also conduct experiments with a Franka Panda to show that our method leads to superior performance on a real robot. It significantly accelerates policy learning for different tasks, achieving success in fewer steps than the baseline. The videos are presented at https://sunlighted.github.io/RRM-web/.