FA-RDP: A Frequency-Adaptive Reactive Diffusion Policy for Contact-Rich Manipulation
2026-07-30 • Robotics
Robotics
AI summaryⓘ
The authors address a problem in robot manipulation where robots need to plan actions differently before and after touching an object. Before contact, many possible movements can work, so keeping options open is important, but after contact, the robot must react quickly to forces. They created FA-RDP, a method that changes how often it updates its plans: it uses slower updates to keep options before touching and faster updates for quick reactions after. They also developed a technique to make sure the robot's planned actions are physically possible. Their tests show this method works well at both maintaining diverse plans and reacting quickly.
contact-rich manipulationdiffusion policymultimodalityforce feedbackTransformersampling frequencyaction manifoldreactive controlManifold Consistency Distillationrobot action planning
Authors
Lifeng Zhuo, Wendi Chen, Han Xue, Shirun Tang, Jun Lv, Cewu Lu, Chuan Wen
Abstract
In contact-rich manipulation, action multimodality and reactivity dominate different stages of a single episode. Before contact, multiple trajectories might be equally valid, making it important to preserve diverse action modes. After contact, geometric constraints and force limits narrow the solution space, while successful execution demands rapid responses to force feedback. However, standard diffusion policies use a fixed inference frequency and sampling steps throughout the episode, forcing a fundamental compromise: low-frequency, multi-step sampling better preserves pre-contact multimodality but responds slowly to force feedback, whereas high-frequency sampling improves reactivity but tends to collapse distinct pre-contact modes. To resolve this tradeoff, we present FA-RDP, a frequency-adaptive reactive diffusion policy. A shared multi-frequency visual-force Transformer predicts action chunks at both low and high frequencies, while a learned multimodality indicator dynamically selects multi-step low-frequency sampling before contact and one-step high-frequency sampling as action ambiguity decreases. We further introduce Manifold Consistency Distillation (MCD), which reparameterizes the diffusion network to predict actions on the robot action manifold while retaining DDPM-based residual supervision. Experiments on three contact-rich manipulation tasks show that FA-RDP achieves the highest success rate while preserving diverse pre-contact trajectory modes. Code and videos are available at https://fa-rdp.github.io.