When Robots Outweigh Objects: How Inertia Mismatch Shapes Robot Policy Learning
When Robots Outweigh Objects: How Inertia Mismatch Shapes Robot Policy Learning
This research investigates how inertia mismatch between heavy robot arms and light objects affects policy learning. Through controlled simulation experiments varying robot and object masses, the authors find that reinforcement learning succeeds across a broad range of mass combinations, while behavior cloning learns more reliable policies when a lighter robot manipulates a heavier object. They also demonstrate that increasing data narrows but does not close the performance gap between methods, and that behavior cloning degrades as simulated robot arm inertia increases via gearing even with constant weight. The work explores practical ways to lower contact inertia by selecting lower-inertia workspaces or steering demonstrators toward lower-inertia task solutions, showing that such demonstrations are imitated more reliably.
The study uses controlled simulation experiments to show that while reinforcement learning can succeed across a wide range of mass combinations, behavior cloning performs more reliably when a lighter robot manipulates a heavier object, and highlights practical strategies to mitigate inertia mismatch without new hardware.
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.
- Environment
- SimulationOpen source S3
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
that even extreme inertia mismatch can be overcome with sufficient data. However, it remains unclear whether high inertia imposes a hidden cost on learning, potentially worsening scaling with data or lowering the performance ceiling of modern robot learning methods. We investigate this question through controlled simulation experiments that vary robot and object masses. We find that reinforcement learning succeeds ac
Open source S3
ore ways to lower contact inertia on the arm one already has, either by choosing a lower-inertia workspace before collecting data or by steering the demonstrator toward lower-inertia solutions of the task. In both simulated cases, the lower-inertia demonstrations are imitated more reliably. Project website: https://younghyopark.me/inertia-mismatch/
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org