TempoVLA: Learning Speed-Controllable Vision-Language-Action Policies

2026-06-04 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Robotics & Autonomous Systems, Artificial Intelligence & Machine Learning · Depth: Expert, medium

Summary

TempoVLA introduces a novel Vision-Language-Action (VLA) model designed for speed-controllable robot manipulation, addressing the limitation of existing VLAs that operate at a single fixed speed. Robot tasks often require alternating between fast transit and slow, precise contact phases. TempoVLA leverages the observation that action magnitude directly influences robot speed, enabling explicit speed conditioning. It comprises two main components: Variable-Speed Trajectory Augmentation (VSTA), which re-times demonstration data to various target speeds while preserving motion semantics, and a model-side mechanism that feeds the desired speed to the policy. Experiments confirm VSTA achieves requested speeds with minimal error and boosts default 1x performance. TempoVLA demonstrates flexible bidirectional speed control in simulation and real-world tasks, and can achieve dynamic speed adjustments for high-risk and low-risk phases when integrated with a large multimodal model.

Key takeaway

For Robotics Engineers developing manipulation policies, TempoVLA offers a critical solution for tasks demanding variable execution speeds. If your applications involve alternating between fast transit and precise contact phases, you should consider integrating speed-controllable Vision-Language-Action models. This approach allows for dynamic speed adjustments, enhancing both safety during high-risk operations and efficiency in low-risk phases, moving beyond fixed-speed limitations. You can leverage data augmentation techniques like VSTA to improve policy performance and adaptability.

Key insights

TempoVLA enables explicit, dynamic speed control for robot manipulation by conditioning Vision-Language-Action models on action magnitude.

Principles

• Action magnitude directly governs robot execution speed.
• Variable-speed trajectory augmentation boosts VLA performance.

Method

TempoVLA combines Variable-Speed Trajectory Augmentation (VSTA) to re-time demonstrations by merging/splitting actions, with a model-side conditioning mechanism that explicitly feeds the target speed to the policy.

In practice

• Implement dynamic speed control for high-risk contact tasks.
• Augment VLA training data with variable-speed trajectories.

Topics

• Vision-Language-Action Models
• Robot Manipulation
• Speed Control
• Trajectory Augmentation
• Robotics Policies
• Multimodal Models

Best for: Research Scientist, AI Scientist, Robotics Engineer, Machine Learning Engineer

Related on AIssential

• TempoVLA: Learning Speed-Controllable Vision-Language-Action Policies — Robot execution speed can be bidirectionally controlled by explicitly scaling predicted action magnitudes within Vision-Language-Action models.
• TempoVLA: Learning Speed-Controllable Vision-Language-Action Policies — Robot execution speed can be dynamically controlled in VLAs by conditioning on desired speed and augmenting training data.
• From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model — BehaviorVLA improves VLA model robustness by learning temporally coherent behavioral representations for generalized manipulation across environments.
• Your Model Already Knows: Attention-Guided Safety Filter for Vision-Language-Action Models — VLA models' attention heads inherently provide real-time target localization, enabling training-free safety filtering against dynamic obstacles.
• Towards Generalizable Robotic Manipulation in Dynamic Environments — Dynamic manipulation in robotics requires specialized datasets and architectures for effective spatiotemporal reasoning.
• vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models — vla.cpp provides a unified, portable C++ runtime for VLA models, enabling efficient, low-memory inference on diverse robotic hardware.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.