Event
UniManip Open-World Robotic Manipulation Framework Released
1 reports · 1 sources
Haichao Liu and other authors have released the UniManip robotic manipulation framework, which uses an Agent Operation Graph (AOG) to unify responsible task orchestration, state tracking, execution verification, and closed-loop failure recovery, and translates high-level semantic intent into executable primitives through a physically grounded scene graph, after which a safety-aware local planner generates collision-free trajectories. The authors state that UniManip achieves state-of-the-art performance on LIBERO-Pro; in real-world experiments, its task success rates are 22.5% and 25.0% higher than leading VLA models and hierarchical baselines, respectively, and it exhibits zero-shot generalization across multiple tasks, embodiments, and dexterous hands. These results are as reported by the authors, and the project page has been open-sourced.
Report timeline
- UniManip: An Open-World Robotic Manipulation Framework Based on Agentic Operational Graphs
UniManip proposes a unified framework centered on the Agentic Operational Graph (AOG) for task orchestration, state tracking, execution verification, and closed-loop failure recovery in open-world robotic manipulation. The method grounds high-level semantic intent into executable primitives through a physically grounded scene graph, and combines a safety-aware local planner to generate collision-free trajectories. It achieves state-of-the-art performance on LIBERO-Pro, and in real-world experiments its task success rate is 22.5% and 25.0% higher than leading VLA models and hierarchical baselines, respectively.
arXiv cs.RO ·