PyRUA-DataFarm uses LLM agent for real-ready robot data
Original title: PyRUA-DataFarm: LLM Agent as a Cost-Efficient Teleoperator for Real-Ready Robot Data
LLM agent as teleoperator rebuilds tasks in simulation and renders photographic video data at $0.80 per episode.
PyRUA-DataFarm employs an LLM agent to take over teleoperator duties for robot demonstrations. The agent receives a data request and one calibration episode, then rebuilds the task in simulation by writing the scene, expert controller, and success judge. A video model conditioned on real images renders each rollout as photographic video with consistent camera views. Independent judges certify all episodes before admission.
- One hour of three-view data costs $107 on average
- Policies trained on farm data achieve 93 successes vs 84 for teleoperation in 100 trials
- Under unseen conditions, farm data keeps 87 successes where teleoperation falls to 46
- Adapting a task costs $21 API usage plus $5 for render contract
- Further episodes cost $0.80 on average
From the original article
1 min read at arxiv.orgTeleoperation remains the default way to collect robot demonstrations, but every episode costs robot and operator time, and every new task starts from scratch. We ask whether a coding agent can take over the teleoperator's job.
By Kewei Zhang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.