AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation

2026-07-01

AI Summary

This work addresses the limitations of existing imitation learning–based visuomotor policies, which typically employ fixed execution speeds and prediction horizons, thereby struggling to adapt to dynamically varying task complexity and compromising both efficiency and flexibility. To overcome this, the authors propose a novel framework for adaptive motion speed learning that requires no explicit phase or speed annotations. The approach formulates speed modulation as a trajectory optimization problem over a set of candidates, leveraging discrete cosine transform (DCT) to enable non-integer frequency-domain scaling of action sequences. This allows rapid execution during simple phases and fine-grained control during complex ones, while dynamically adjusting the effective prediction horizon. Integrated with model-free policy optimization and a composite cost–based trajectory selection mechanism, the method significantly reduces task completion time while improving success rates, with learned speeds aligning closely with semantic task phases.

Abstract

Different stages of manipulation tasks exhibit varying levels of difficulty, suggesting stage-dependent motion speeds and temporal prediction horizons. However, existing IL-based visuomotor policies typically imitate the execution speed of expert demonstrations and operate with a fixed temporal prediction horizon, limiting flexibility and overall task throughput. In this paper, we introduce AutoSpeed, a model-agnostic learning framework that enables existing visuomotor policies to predict trajectories with stage-adaptive motion speeds, without requiring speed or stage annotations. We treat future trajectories at different speeds as candidate optimization targets, evaluate each candidate using a composite cost that trades off prediction error against prediction horizon, and optimize the policy toward the minimum-cost candidate. With a fixed-length action sequence, speed modulation adjusts the effective temporal prediction horizon: simple stages are executed faster with a longer prediction horizon, whereas complex stages are executed more slowly with a shorter prediction horizon. Specifically, we implement speed modulation in the frequency domain via the discrete cosine transform (DCT), which enables smooth, non-integer speed scaling and thus preserves motion continuity. Extensive evaluations show that AutoSpeed substantially reduces task execution time while also improving success rates. Under the AutoSpeed framework, the inferred motion speeds exhibit a strong correspondence with task stages.

Problem

• imitation learning
• visuomotor policy
• motion speed
• temporal prediction horizon
• robot manipulation

Innovation

• stage-adaptive motion speed
• annotation-free learning
• temporal prediction horizon
• discrete cosine transform (DCT)
• visuomotor policy