SpeedTuning: Speeding Up Policy Execution with Lightweight Reinforcement Learning

2026-08-10

AI Summary

This work addresses the limitation of existing robotic imitation learning policies, which often suffer from slow execution speeds and lack effective mechanisms for acceleration without collecting new data. To overcome this, we propose SpeedTuning, a novel framework that dynamically predicts optimal action execution speeds through lightweight reinforcement learning, without modifying the base policy or requiring additional demonstrations. SpeedTuning introduces an adaptive speed modulation mechanism that combines temporal action scaling with unsupervised optimization, revealing the inherent trade-off between execution speed and task success. Evaluated on tasks such as pouring, throwing, and grasping, our method achieves over 2.4× speedup while maintaining success rates comparable to the original policy, significantly outperforming baseline approaches like fixed-rate linear interpolation.

Abstract

While learned robotic policies hold promise for advancing generalizable manipulation, their practical deployment is often hindered by suboptimal execution speeds. Imitation learning policies are inherently limited by hardware constraints and the speed of the operator during data collection. In addition, there are no established methods for accelerating policies learned via imitation, and the empirical relationship between execution speed and task success remains underexplored. To address these issues, we introduce SpeedTuning, a reinforcement learning framework specifically designed to enhance the speed of manipulation policies. SpeedTuning learns to predict the optimal execution speed for actions, thereby complementing a base policy without necessitating additional data collection. We provide empirical evidence that SpeedTuning achieves substantial improvements in execution speed, exceeding 2.4x speed-up, while preserving an adequate success rate compared to both the original task policy and straightforward speed-up methods such as linear interpolation at a fixed speed. We evaluate our approach across a diverse set of dynamic and precise tasks, including pouring, throwing, and picking, demonstrating its effectiveness and robustness in enhancing real-world robotic manipulation.

Problem

• policy execution speed
• imitation learning
• robotic manipulation
• task success rate
• execution acceleration

Innovation

• SpeedTuning
• reinforcement learning
• policy acceleration
• robotic manipulation
• imitation learning