The Unexpired Plan: A Free Monitor for Accelerated Diffusion Policies

2026-10-05

AI Summary

This study addresses the lack of low-cost, training-free monitoring mechanisms for diffusion policy acceleration, which hinders the safe reuse of computational results. We propose a training-free monitoring framework based on "unexpired plans" that pioneers using the previous planning trajectory as a zero-cost reference signal. Rather than measuring feature distances, our method authenticates accelerated steps by evaluating action deviations and reformulates the monitor as a combination of statistics and response policies, effectively circumventing traditional gating forgetting issues. Experiments demonstrate that this approach achieves 1.55× to 3.09× single-step computation reduction across four policy families while strictly maintaining predefined precision margins, significantly outperforming existing accelerator configurations.

Abstract

Training-free acceleration of a diffusion policy is accepted when an internal similarity signal reports that the shortcut changed nothing. We price every monitor a control loop can afford in closed-loop success rather than feature distance, over $114$ accelerator configurations and four policy families: two accelerators each clear their own gate's bar and log every reuse as certified, while on one task one finishes every episode and the other none. A monitor is two designs, not one --- the statistic it reads, and what it does when that statistic fires. Published gates re-arm after every rejection, and under that response even an oracle handed every forward pass and the exact local action error loses twenty points; absorbing the same statistic costs one point, and most of its speed. What makes a response that never forgets affordable is a statistic that rarely fires, and what it must measure is deviation from the policy the accelerator replaced --- which a chunked policy has already paid for, its last plan not yet expired and free to read. Guarding every call this way cuts per-call compute by $1.55$--$3.09\times$, where any reference-requiring check at the same coverage would have to stop accelerating altogether. On three of our four families the schedule alone already holds the pre-stated $\pm2$-point margin. What the monitor is measurably worth shows in three places: on the fourth family, where it rescues the candidate selection landed on; on four configurations it did not select; and on a contact-rich fifth family, chosen where the schedule was expected to fail and run after every design choice was frozen, where no unmonitored arm at its speed holds the margin and the monitored one does.

Problem

• diffusion policy
• training-free acceleration
• monitor
• closed-loop control
• action chunking

Innovation

• Training-free acceleration
• Diffusion policy
• Monitor design
• Chunked policy
• Closed-loop evaluation