Control-Aware Command Generation for Asynchronous Control and Action Inference with Vision-Based Policies towards Fast Execution
Organizations: Center for Advanced Intelligence Project RIKEN · Graduate School of Engineering The University of Tokyo
Abstract
Vision-based control policies, including vision-language-action models, have enabled robots to perform increasingly complex tasks. However, their execution speed is often limited by the high computational cost of action inference. In this work, we investigate what can be achieved at the control level without modifying the underlying policy learning process. We introduce an asynchronous inference and control framework that enables fast task execution while maintaining smooth and stable robot motions. By formulating the generation of continuous target positions from discontinuous model predictions as a tracking control problem, our method makes asynchronous inference practical for real robots. We further accelerate execution through adaptive action skipping based on the variance of predicted positions. Experiments with Action Chunking Transformer and OpenVLA-OFT on a bimanual robot demonstrate substantial speedups while preserving high task success rates.
Figures & tables
| Task | ACT +Sync | ACT +Async | ACT +CAC +Async | ACT +CAC +Async +Skip | ACT +CAC +Async +Adaptive |
|---|---|---|---|---|---|
| Box handover | 100.0% | 100.0% | 100.0% | 86.7 % | 96.7% |
| Case stacking | 100.0% | 100.0% | 100.0% | 100.0% | 100.0% |
| Pen storage | 30.0% | 63.3% | 70.0% | 0 % | 83.3% |
| Task | ACT +Sync | ACT +Async | ACT +CAC +Async | ACT +CAC +Async +Skip | ACT +CAC +Async +Adaptive |
| Box handover | 16.2 0.3 | 11.0 0.2 | 11.1 0.1 | 7.2 0.3 | 8.6 0.1 |
| Case stacking | 34.4 1.4 | 21.8 0.2 | 22.2 0.6 | 12.9 0.3 | 19.2 0.6 |
| Pen storage | 52.6 0.9 | 33.4 1.2 | 38.6 1.0 | - | 34.4 0.8 |
| All values are in seconds. The mean and standard deviation are shown. | |||||
| Task | OpenVLA-OFT +Sync | OpenVLA-OFT +CAC +Sync | OpenVLA-OFT +CAC +Async |
|---|---|---|---|
| success rate | 8/10 | 9/10 | 8/10 |
| episode duration [s] | 68.2 0.1 | 68.2 0.1 | 59.0 0.1 |
| task execution time [s] | 49.0 0.2 | 52.6 1.6 | 44.7 2.5 |
Robotic Control
Robotic control studies how machines use sensing and computation to produce reliable movement and interaction in changing environments.
See all 620 papers