LIBERO-Dynamic: Benchmarking Robotic Manipulation in Dynamic Environments
Most simulation benchmarks for robotic manipulation keep target objects static and treat policy inference as instantaneous.
ProofPaper ↗
Key points
- These assumptions simplify evaluation but hide how a moving target evolves during computation and how stale an observation is when an action is executed.
- We introduce LIBERO-Dynamic, a dynamic extension of all 40 LIBERO tasks.
- We evaluate six pretrained policies on 1,800 dynamic conditions per configuration, compare execution strategies and action horizons with fixed checkpoints, and fine-tune three architectures using dynamic and mixed data.
- LIBERO-Dynamic makes object motion, inference latency, and action execution explicit and controllable for evaluating and improving robot policies in moving-target manipulation.
Sources (1)
- [1]LIBERO-Dynamic: Benchmarking Robotic Manipulation in Dynamic EnvironmentsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 17, 09:17 AM
Most simulation benchmarks for robotic manipulation keep target objects static and treat policy inference as instantaneous.
These assumptions simplify evaluation but hide how a moving target evolves during computation and how stale an observation is when an action is executed.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 17, 2026GazeAct: A Human-Gaze-Annotated Manipulation Dataset for Attention-Regularized Policies
- Oct 16, 2026Look Before You Act: Bridging Active Perception and Vision-Language-Action Models via Observation-Aligned Data Adaptation
- Oct 16, 2026DLite: History-free Speculative Decoding for Agentic AI
- Oct 15, 2026Embodiment-Classifier Guidance: Transferring Frozen VLAs Across Robot Dynamics without Demonstrations
- Oct 15, 2026TS-WAM: Temporal Latent Actions and World-Action Coupling for Vision-Language Robot Control
- Oct 15, 2026What Should Robot Policies Learn to Predict? Rethinking Supervision Targets in Vision-Language-Action Learning