RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory
RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory
This research introduces RoboArchitect, a system that automates the design of robotic agent harnesses by observing robot rollouts, diagnosing failures, and iteratively revising control strategies. Without updating any model weights, RoboArchitect increases average success rates from 9.7% to 43.3% on unseen scenes and from 12.7% to 40.2% on unseen tasks across multiple benchmarks. The system automatically designs intuitive control strategies, such as lifting a jammed arm or navigating around obstacles, and evolves its own persistent memory to carry learned experience into later decisions. These results suggest that substantial improvements in robotic manipulation can be achieved by optimizing the agent harness rather than fine-tuning model weights.
The study demonstrates that automating the design of robotic agent harnesses using a coding model can significantly improve performance on unseen scenes and tasks without updating model weights, achieving up to a 4.5x increase in success rates across multiple benchmarks.
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
improvement factor
4.5
View original evidence
ing on a novel setup. In RoboArchitect, a coding model observes a robot's rollouts, diagnoses failures, and iteratively revises the harness around the frozen policy. The harness evolves on a set of episodes for its optimization and is evaluated on held-out scenes and tasks. Without updating any model weights, RoboArchitect raises average success from 9.7% to 43.3% on unseen scenes (4.5x) and from 12.7% to 40.2% on un
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ing on a novel setup. In RoboArchitect, a coding model observes a robot's rollouts, diagnoses failures, and iteratively revises the harness around the frozen policy. The harness evolves on a set of episodes for its optimization and is evaluated on held-out scenes and tasks. Without updating any model weights, RoboArchitect raises average success from 9.7% to 43.3% on unseen scenes (4.5x) and from 12.7% to 40.2% on un
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org