Reparameterization Flow Policy Optimization

TL;DR

This paper introduces Reparameterization Flow Policy Optimization (RFO), a novel method that leverages flow policies within the RPG framework to improve sample efficiency and performance in reinforcement learning tasks involving complex dynamics.

Contribution

The paper establishes the connection between flow policies and RPG, proposes RFO with regularization for stability and exploration, and demonstrates its effectiveness on diverse tasks.

Findings

RFO achieves nearly 2x the reward of the baseline on a soft-body quadruped task.

Flow policies can be effectively integrated into RPG without intractable likelihood computations.

RFO outperforms existing methods in various locomotion and manipulation tasks.

Abstract

Reparameterization Policy Gradient (RPG) has emerged as a powerful paradigm for model-based reinforcement learning, enabling high sample efficiency by backpropagating gradients through differentiable dynamics. However, prior RPG approaches have been predominantly restricted to Gaussian policies, limiting their performance and failing to leverage recent advances in generative models. In this work, we identify that flow policies, which generate actions via differentiable ODE integration, naturally align with the RPG framework, a connection not established in prior work. However, naively exploiting this synergy proves ineffective, often suffering from training instability and a lack of exploration. We propose Reparameterization Flow Policy Optimization (RFO). RFO computes policy gradients by backpropagating jointly through the flow generation process and system dynamics, unlocking high sample efficiency without requiring intractable log-likelihood calculations. RFO includes two tailored regularization terms for stability and exploration. We also propose a variant of RFO with action chunking. Extensive experiments on diverse locomotion and manipulation tasks, involving both rigid and soft bodies with state or visual inputs, demonstrate the effectiveness of RFO. Notably, on a challenging locomotion task controlling a soft-body quadruped, RFO achieves almost $2\times$ the reward of the state-of-the-art baseline.