MotionForge benchmark tests robot manipulation in changing dynamic scenes
MotionForge: A Data Generation Pipeline and Large-Scale Benchmark for Long-Horizon Manipulation of Dynamic Objects with Domain Shifts
Robotics
Summary
Robots learning to handle objects usually do so in still or nearly still settings. The authors created MotionForge, a tool that simulates more realistic situations where objects keep moving and the environment keeps changing. It also tests how well robot programs work when things like background, lighting, or object speed suddenly change. Their tests show current robot skills struggle with these challenges, revealing what needs improvement for robots to work better in real-world changing conditions.
What this means in practice
- •For industrial automation teams: Evaluate and improve robot programs for handling moving parts in complex factory settings where lighting and object placement vary.
- •For robotics software developers: Test robotic control algorithms for adaptability to unexpected changes in the environment over extended operation periods.
Authors
Mohan Liu, Dengchen Mei, Haotian Xian, Ruyang Han, Jiayi Sun, Xuanyu Chen, Haitian Zhang, Luxi Li, Kaimin Mao, Lin Wang
Abstract
Recent advances in learning-based robot policies have demonstrated promising progress, yet they are predom- inantly evaluated in static or quasi-static environments. In dynamic manipulation, objects and scenes continuously evolve while the robot perceives, reasons, and acts. However, recent dynamic simulation benchmarks largely focus on short-horizon, reactive interactions with simple motion patterns and offer limited support for both systematic evaluation under domain shifts and model-agnostic real-time execution protocols. To bridge these gaps, we introduce MotionForge, the first large- scale simulation benchmark and data-generation pipeline tailored to jointly evaluate domain shifts and long-horizon interaction in dynamic manipulation. MotionForge comprises 40 dynamic interaction tasks spanning 11 distinct motion patterns, with dedicated support for 17 long-horizon tasks. Our benchmark introduces two key novelties: (1) a systematic evaluation protocol for assessing policy robustness under both single-factor (e.g., only backgrounds shift) and joint domain shifts (e.g., simultaneous shifts of objects, backgrounds, lighting, and speed); and (2) a decoupled, latency-aware execution protocol where the environ- ment continuously evolves independently of policy inference time. Extensive evaluations of representative general-purpose robot policies on our benchmark reveal substantial limitations under joint domain shifts. These findings expose a critical gap between current policy capabilities and the requirements of robust long- horizon manipulation of dynamic objects under domain shifts, establishing MotionForge as a comprehensive testbed for future research in embodied AI.