HDFlow: Hierarchical Diffusion-Flow Planning for Long-horizon Tasks

TL;DR

HDFlow introduces a hierarchical planning framework combining diffusion and rectified flow models to improve long-horizon task planning, enabling efficient and effective behavior generation in complex tasks.

Contribution

The paper presents HDFlow, a novel hierarchical diffusion-flow framework that enhances long-horizon planning by combining high-level diffusion planning with low-level rectified flow trajectory generation.

Findings

HDFlow outperforms state-of-the-art methods in furniture assembly tasks.

HDFlow demonstrates strong generalization on diverse locomotion and manipulation benchmarks.

The approach enables real-time, dense trajectory generation for complex tasks.

Abstract

Recent advances in generative models have shown promise in generating behavior plans for long-horizon, sparse reward tasks. While these approaches have achieved promising results, they often lack a principled framework for hierarchical decomposition and struggle with the computational demands of real-time execution, due to their iterative denoising process. In this work, we introduce Hierarchical Diffusion-Flow (HDFlow), a novel hierarchical planning framework that optimally leverages the strengths of diffusion and rectified flow models to overcome the limitations of single-paradigm generative planners. HDFlow employs a high-level diffusion planner to generate sequences of strategic subgoals in a learned latent space, capitalizing on diffusion's powerful exploratory capabilities. These subgoals then guide a low-level rectified flow planner that generates smooth and dense trajectories, exploiting the speed and efficiency of ordinary differential equation (ODE)-based trajectory generation. We evaluate HDFlow on four challenging furniture assembly tasks in both simulation and real-world, where it significantly outperforms state-of-the-art methods. Furthermore, we also showcase our method's generalizability on two long-horizon benchmarks comprising diverse locomotion and manipulation tasks. Project website: https://hdflow-page.github.io/