Efficient Online RL Fine Tuning with Offline Pre-trained Policy Only

TL;DR

This paper introduces PORL, a method for efficient online reinforcement learning fine-tuning that relies solely on offline pre-trained policies, avoiding the need for pre-trained Q-functions and enabling direct fine-tuning of behavior cloning policies.

Contribution

The paper presents PORL, a novel approach that initializes Q-functions from scratch during online fine-tuning, improving stability and applicability without requiring offline Q-functions.

Findings

Achieves competitive performance with existing offline-to-online RL methods.

Enables direct fine-tuning of behavior cloning policies.

Eliminates the need for offline pre-trained Q-functions.

Abstract

Improving the performance of pre-trained policies through online reinforcement learning (RL) is a critical yet challenging topic. Existing online RL fine-tuning methods require continued training with offline pretrained Q-functions for stability and performance. However, these offline pretrained Q-functions commonly underestimate state-action pairs beyond the offline dataset due to the conservatism in most offline RL methods, which hinders further exploration when transitioning from the offline to the online setting. Additionally, this requirement limits their applicability in scenarios where only pre-trained policies are available but pre-trained Q-functions are absent, such as in imitation learning (IL) pre-training. To address these challenges, we propose a method for efficient online RL fine-tuning using solely the offline pre-trained policy, eliminating reliance on pre-trained Q-functions. We introduce PORL (Policy-Only Reinforcement Learning Fine-Tuning), which rapidly initializes the Q-function from scratch during the online phase to avoid detrimental pessimism. Our method not only achieves competitive performance with advanced offline-to-online RL algorithms and online RL approaches that leverage data or policies prior, but also pioneers a new path for directly fine-tuning behavior cloning (BC) policies.