Robo-MLT: Dual-System Robotic Manipulation for Long-Horizon Medical Laboratory Tasks
Robo-MLT: Dual-System Robotic Manipulation for Long-Horizon Medical Laboratory Tasks
This research introduces Robo-MLT, a framework that couples explicit workflow monitoring with learned action generation to handle long-horizon medical laboratory tasks. The system uses a vision-language planner to compose atomic skill sequences and a multiframe execution monitor to detect task completion or infeasibility, triggering replanning when necessary. Evaluated on blood gas analysis and sample tube sorting with 50 nominal configurations per method and 20 sample-removal configurations, Robo-MLT achieves 82% and 70% task success rates respectively, outperforming baseline policies by 30 and 22 percentage points. The system demonstrates the ability to revise plans after operator intervention, such as removing a handled sample, indicating improved reliability and recoverability for complex laboratory workflows.
The study evaluates a dual-system framework combining explicit workflow monitoring with learned action generation on specific medical laboratory tasks, reporting task success rates of 82% and 70% against baselines under tested conditions involving sample removal.
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
percentage points improvement over baseline
30
View original evidence
ime coordinates it with the monitor. We evaluate Robo-MLT against four policy baselines and a VLM-based monitor baseline on blood gas analysis and blood sample tube sorting, with 50 nominal configurations per method and 20 sample-removal configurations. Robo-MLT achieves task success rates of 82\% and 70\%, exceeding $\pi_{0.5}$ by 30 and 22 percentage points, respectively. After an operator removes the sample being
Open source S4percentage points improvement over baseline
22
View original evidence
ime coordinates it with the monitor. We evaluate Robo-MLT against four policy baselines and a VLM-based monitor baseline on blood gas analysis and blood sample tube sorting, with 50 nominal configurations per method and 20 sample-removal configurations. Robo-MLT achieves task success rates of 82\% and 70\%, exceeding $\pi_{0.5}$ by 30 and 22 percentage points, respectively. After an operator removes the sample being
Open source S4nominal configurations per method
50
View original evidence
ime coordinates it with the monitor. We evaluate Robo-MLT against four policy baselines and a VLM-based monitor baseline on blood gas analysis and blood sample tube sorting, with 50 nominal configurations per method and 20 sample-removal configurations. Robo-MLT achieves task success rates of 82\% and 70\%, exceeding $\pi_{0.5}$ by 30 and 22 percentage points, respectively. After an operator removes the sample being
Open source S4sample-removal configurations
20
View original evidence
ime coordinates it with the monitor. We evaluate Robo-MLT against four policy baselines and a VLM-based monitor baseline on blood gas analysis and blood sample tube sorting, with 50 nominal configurations per method and 20 sample-removal configurations. Robo-MLT achieves task success rates of 82\% and 70\%, exceeding $\pi_{0.5}$ by 30 and 22 percentage points, respectively. After an operator removes the sample being
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ime coordinates it with the monitor. We evaluate Robo-MLT against four policy baselines and a VLM-based monitor baseline on blood gas analysis and blood sample tube sorting, with 50 nominal configurations per method and 20 sample-removal configurations. Robo-MLT achieves task success rates of 82\% and 70\%, exceeding $\pi_{0.5}$ by 30 and 22 percentage points, respectively. After an operator removes the sample being
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org