FreeSpeed: Training-Free Speed Control for Generative Robot Policies

2026-10-05

AI Summary

This study addresses the limitations of imitation learning policies, which typically execute at fixed speeds, resist online adjustment, and are prone to out-of-distribution failures. To overcome these challenges, this work proposes a training-free action chunk post-processing module. By leveraging action direction inconsistency as a task-criticality signal, the method adaptively determines speed adjustment boundaries and resamples and scales action chunks generated by pretrained policies, enabling flexible execution speed control without retraining. Experimental results demonstrate that the proposed module supports variable execution speeds ranging from 0.22× to 2.53× in simulation. In real-world settings, it achieves an average success rate of 94.0%, matching the original policy performance, while enabling controllable execution speeds between 0.38× and 1.97×.

Abstract

Online control of execution speed is essential for deploying robot policies in real-world scenarios, as robots may need to speed up under time constraints or slow down to facilitate human interaction and improve safety. However, imitation-learned policies inherit the execution speed of their demonstrations, and test-time speed modification can introduce unrecoverable out-of-distribution observations, reducing task success. We observe that the directional inconsistency of action chunks reflects task-phase criticality, indicating how aggressively action step lengths can be modified while preserving task success. Based on this observation, we introduce FreeSpeed, a training-free module that post-processes action chunks from pretrained policies. FreeSpeed resamples each predicted chunk at the requested rate, then uses directional inconsistency between adjacent actions as the primary signal for rescaling. This signal adaptively determines how closely the execution speed can approach the requested speed, allowing flexible speed adjustment within the evaluated limits without compromising task success. Across three policy families and 50 simulated tasks, FreeSpeed supports online speed changes, with realized execution rates spanning 0.22x to 2.53x among settings that preserve per-task success. Across four real-world manipulation tasks, FreeSpeed achieves an average success rate of 94.0%, matching the frozen policy's 93.8%, while realizing execution rates from 0.38x to 1.97x.

Problem

• speed control
• generative robot policies
• imitation learning
• out-of-distribution
• online execution

Innovation

• Training-Free
• Speed Control
• Action Chunks
• Directional Inconsistency
• Generative Robot Policies