Geometric pretraining boosts robot manipulation learning efficiency
Improving Imitation Learning Efficiency for Manipulation through Geometric Prior Pretraining
Robotics
Summary
Teaching robots to do new tasks usually needs lots of examples and retraining, which can be slow and costly. The authors show that by first training robots on simple scenes showing basic shapes like a hand, an object, and a flat surface, the robots learn important spatial relationships. This geometric prior helps them learn new tasks faster with fewer examples, even across different robot types and in real-world settings. Their approach doesn’t need complicated visuals or robot-specific details, making it more flexible and efficient.
What this means in practice
- •For robotics engineers: Improve early training success rates for robot manipulation tasks with minimal demonstration data by using geometric pretraining.
- •For automation system builders: Accelerate fine-tuning of manipulation policies across different robot morphologies using a shared geometric prior.
Authors
Shogo Iwakata, Tomohiro Motoda, Ryosuke Yamada, Koshi Makihara, Ryoichi Nakajo, Keitaro Tanaka, Masaki Murooka, Roman Mykhailyshyn, Hirokatsu Kataoka, Shigeo Morishima, Yukiyasu Domae
Abstract
Applying an imitation learning policy to a new manipulation task usually requires collecting new demonstrations and retraining the model, which makes sample efficiency a practical concern. Pretraining on large-scale robot datasets is effective in this respect, but such datasets are costly to collect and train on, while data augmentation techniques typically require a new round of data generation and retraining for each task. A complementary question is what useful prior can be provided to a policy at negligible cost before any task-specific data are collected. In this study, we construct a geometric visual pretraining dataset in which each scene contains only a plane, an object, and a hand, and trajectories are generated automatically. The scenes contain neither textures nor backgrounds; pretraining primarily exposes the policy to the geometric relationship between the hand and the object. Furthermore, representing the hand as a cube avoids tailoring the dataset to a specific robot morphology. We evaluate this geometric prior using ACT on three simulated robots across five manipulation tasks each, as well as on three real-world robot tasks. Across many of these robot--task combinations, fine-tuning from the geometric prior achieves higher success rates in the early stages of training than training from scratch while using only a small number of task demonstrations. These results suggest that even highly simplified geometric scenes can provide a useful initialization that transfers across robots and to real-world tasks when task data are limited.