Researchers and developers are making significant strides in the field of robotics through the application of foundation models and machine learning frameworks. These advancements enable robots to self-learn their morphology, kinematics, and topology, as well as perform dexterous manipulation tasks. The use of generalist robot policies and the release of a new robot foundation model, Octo, are expected to become standard building blocks in robotics, supporting language and goals while efficiently finetuning on various robots.
[RO] Foundation Models in Robotics: Applications, Challenges, and the Future https://t.co/dCMrzmwp0k This paper presents an overview of the applications of pretrained foundation models in robotics. Traditional deep learning models in robotics are trained on small datasets… https://t.co/ylz6OdvYlH
Generalist robot policies, like GNM, RT-X, etc., will be a standard building block in robotics. We released a new robot foundation model, Octo, designed to run on many robots, supports language and goals, and finetunes efficiently! Check it out: https://t.co/62rjh576sd A 🧵👇 https://t.co/K636JS4Npb
MIT researchers discuss their vision-based controller, which helps robots w/in-hand reorientation for dexterous manipulation tasks. The controller can move novel & complex objects to their target orientation to assist w/tool use: https://t.co/6FWDCXpMeu https://t.co/zdZXc498Tq
Foundation Models in Robotics: Applications, Challenges, and the Future paper page: https://t.co/nEkNL12cDV We survey applications of pretrained foundation models in robotics. Traditional deep learning models in robotics are trained on small datasets tailored for specific… https://t.co/qnAG7jHrxr https://t.co/0nE7rgxUeD
A new #MachineLearning framework enables robots to self-learn their morphology, kinematics, and topology. Learn more in Science Robotics: https://t.co/txHBXJYybZ https://t.co/6dZbYHYJoi