- Robots struggle to learn from each other and rely on human instruction
- New UC Berkeley research shows process could be automated
- This would eliminate the difficulties of manually training robots.
Although robots are increasingly integrated into real-world environments, one of the major challenges in robotics research is ensuring that devices can effectively adapt to new tasks and environments.
Traditionally, training to master specific skills requires large amounts of data and specialized training for each robot model. But to overcome these limitations, researchers are now focusing on creating computational frameworks that enable the transfer of skills between different robots.
A new development in robotics comes from researchers at UC Berkeley, who introduced RoVi-Aug, a framework designed to augment robotics data and facilitate skills transfer.
The challenge of transferring skills between robots
To facilitate the robotics training process, it is necessary to be able to transfer acquired skills from one robot to another, even if these robots have different hardware and design. This capability would make it easier to deploy robots in a wide range of applications without having to retrain them from scratch.
However, in many current robotics datasets, there is an uneven distribution of scenes and demonstrations. Some robots, such as the Franka and xArm manipulators, dominate these datasets, making it more difficult to generalize learned skills to other robots.
To address the limitations of existing datasets and models, the UC Berkeley team developed the RoVi-Aug framework that uses state-of-the-art diffusion models to augment robotic data. The framework works by producing synthetic visual demonstrations that vary in both robot type and camera angles. This allows researchers to train robots on a wider range of demonstrations, enabling more effective skills transfer.
The framework consists of two key elements: the robot augmentation module (Ro-Aug) and the viewpoint augmentation module (Vi-Aug).
The Ro-Aug module generates demonstrations involving different robotic systems, while the Vi-Aug module creates demonstrations captured from different camera angles. Together, these modules provide a richer and more diverse data set for training robots, helping to bridge the gap between different models and tasks.
“The success of modern machine learning systems, particularly generative models, demonstrates impressive generalizability and motivates robotics researchers to explore how to achieve similar generalizability in robotics,” Lawrence Chen (PhD candidate, AUTOLab, EECS & IEOR, BAIR, UC Berkeley) and Chenfeng Xu (PhD candidate, Pallas Lab & MSC Lab, EECS & ME, BAIR, UC Berkeley), said Xplore Technology.