Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience
Evo-RL: Iterative Robot Reinforcement Learning from Selective Experience
Evo-RL is an iterative reinforcement learning method that integrates human corrections with advantage-guided experience selection to refine robot policies. The approach leverages experience from successful trajectories to improve policy performance, achieving a 30.46% increase in task success rate compared to RECAP on a high-precision manipulation task. The RW-RL-Dataset, spanning over 1,000 hours across four robot embodiments, is made publicly available along with code and model checkpoints.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported duration
>1,000 hours
View original evidence
ive reinforcement learning method that combines human interventions with advantage-guided experience selection. Evo-RL uses human corrections and experience with positive estimated advantages from successful trajectories to iteratively refine the policy. To support learning from these different experience sources, we introduce RW-RL-Dataset, which spans over 1,000 hours across four robot embodiments. On a real-world
Open source S3robot embodiments
4 robots
View original evidence
ive reinforcement learning method that combines human interventions with advantage-guided experience selection. Evo-RL uses human corrections and experience with positive estimated advantages from successful trajectories to iteratively refine the policy. To support learning from these different experience sources, we introduce RW-RL-Dataset, which spans over 1,000 hours across four robot embodiments. On a real-world
Open source S3
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ive reinforcement learning method that combines human interventions with advantage-guided experience selection. Evo-RL uses human corrections and experience with positive estimated advantages from successful trajectories to iteratively refine the policy. To support learning from these different experience sources, we introduce RW-RL-Dataset, which spans over 1,000 hours across four robot embodiments. On a real-world
Open source S3
high-precision manipulation task, Evo-RL improves attempt success rate over RECAP by 30.46\%. Our code, model checkpoints, and dataset are publicly available.
Open source S4
arXiv:2610.13345v1 Announce Type: new Abstract: Existing real-world reinforcement learning methods improve robot policies by combining human corrections with autonomous rollouts. However, accumulated experience varies in its value for policy improvement, and even successful trajectories can contain inefficient or redundant behavior, motivating more selective experience reuse. To this end, we propose Evo-RL, an iterat
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org