Smoother Action Chunking Flow Policy via Prior-Corrected Orthogonal Trust-Region Guidance

2026-05-23

AI Summary

This work addresses the boundary discontinuities in action chunking inference that commonly arise with flow-matching-based robotic policies, which degrade control smoothness. To mitigate this issue, the authors propose a novel approach integrating data-driven prior-corrected guidance weights with an orthogonal trust-region constraint. Specifically, the method enhances correction signals at intermediate timesteps using a learned prior and decomposes the guidance vector into parallel and perpendicular components relative to the current trajectory. A trust-region constraint is then applied exclusively to the perpendicular component to suppress lateral perturbations. This strategy significantly improves action continuity and stability, achieving state-of-the-art performance on the LIBERO benchmark among real-time control (RTC) methods—demonstrating higher task success rates alongside markedly reduced boundary discontinuities, acceleration, and jerk.

Abstract

Flow-matching robot policies commonly use action-chunking inference for efficient closed-loop control, but chunk boundaries can introduce discontinuous action transitions. Existing RTC guidance improves continuity by injecting correction signals during denoising, yet its weight schedule is weak at intermediate timesteps and its unconstrained correction direction may introduce transverse perturbations. We propose POTR, a **p**rior-corrected **o**rthogonal **t**rust-**r**egion guidance method. First, we incorporate a data-prior scale $σ_d$ into the RTC guidance weight, yielding stronger intermediate-time correction. Second, we decompose the guidance vector into components parallel and perpendicular to the denoising velocity, and constrain the perpendicular component within a trust region. On LIBERO with $π_{0.5}$, POTR improves success rate and consistently reduces chunk-boundary discontinuity, acceleration, and jerk compared with RTC. Ablations show that the prior-corrected weight provides the main correction gain, while the orthogonal trust region further improves stability.

Problem

• action chunking
• flow matching
• discontinuous transitions
• robot policy
• motion smoothness

Innovation

• flow matching
• action chunking
• trust-region guidance
• orthogonal decomposition
• robot policy