Acro improves robot task success by choosing when and where to restart
ACRO: Actor-Critic Rollout Orchestration
Robotics
Summary
Robots that follow vision and language instructions sometimes get stuck or make mistakes during tasks. The authors created ACRO, a system that helps decide the best moments to stop and try again from a better spot along the robot’s path. ACRO uses a learned model to predict which restarts might be more successful and physically possible, improving overall task completion rates. This approach was tested both in simulation and on real robots, showing higher success without extra execution time.
What this means in practice
- •For robotics engineers: Improve robot manipulation success by integrating ACRO to decide when to retry actions during task execution based on success predictions.
- •For autonomous system developers: Enhance real-world autonomous agents by using ACRO to orchestrate action retries for higher reliability under limited operation time.
Authors
Junhoo Lee, Seungyeon Kim, Baekseung Kim, Minkyu Kim, Suhyun Jeon, Jimyeong Kim, Nojun Kwak
Abstract
Executing a vision-language-action (VLA) policy involves deciding when to continue an attempt and when to intervene to improve its chances of success. We introduce ACRO, an Actor--Critic rollout orchestration layer that makes these decisions for a fixed VLA. ACRO seeks interventions that offer a more promising continuation, are physically feasible, and leave sufficient time to complete the task. A Critic learned from policy rollouts estimates success prospects to decide when to intervene and select a re-entry target along the executed trajectory. Trajectory Retraction shortcuts the visited path to return the robot to that target, where the same VLA resumes execution from a fresh observation. We evaluate ACRO with GR00T N1.7 and $π_{0.5}$ across SimplerEnv, RoboCasa, and real-robot manipulation. Under the same physical execution budget, including all return motion, ACRO improves success by 10.4, 7.3, and 21.7 percentage points, respectively. It also completes additional tasks as the execution budget grows. These results show that controlling when and where a policy resumes can turn its existing capabilities into more successful task executions. Project page: https://junhoo.me/acro