Dobb·E
About Dobb·E
Dobb·E is a pioneering platform designed to revolutionize home robotics by enabling robots to learn household tasks using a simple imitation learning framework. Targeting researchers and hobbyists, this innovative system leverages the unique demonstration tool called the Stick to make robotic learning more accessible and effective.
Dobb·E offers a free, open-source platform for anyone interested in household robotics. Users can access comprehensive resources, including datasets and models, at no cost. Upgrading to premium resources provides enhanced capabilities for advanced users, fostering innovation in robotic task learning and application development.
Dobb·E boasts a user-friendly interface designed for simplicity and efficiency, ensuring a seamless experience for users of all skill levels. Its intuitive layout facilitates easy navigation through various features, making it accessible for researchers and robotics enthusiasts eager to develop household robotic solutions.
How Dobb·E works
Users interact with Dobb·E by first onboarding and familiarizing themselves with the platform’s core features. They can easily collect demonstration data using the Stick, which captures real-time tasks in household environments. After gathering a few minutes of demonstrations, users utilize the trained models to adapt their robots, enabling swift learning of new tasks in unfamiliar settings.
Key Features for Dobb·E
Imitation Learning Framework
The imitation learning framework of Dobb·E allows robots to learn household tasks through a unique demonstration tool named the Stick. This innovative approach significantly reduces training time, enabling robots to adapt quickly to new environments and tasks, thus transforming the future of home robotics.
Home Pretrained Representations (HPR)
Dobb·E's Home Pretrained Representations (HPR) offer a powerful model pretrained on diverse household tasks. This model streamlines the learning process, providing a robust foundation for robots to quickly adapt to new tasks, enhancing efficiency and effectiveness in various home settings.
Datasets for Training
Dobb·E provides access to an extensive dataset called Homes of New York (HoNY), which contains rich interactions recorded in diverse household environments. This wealth of data empowers users to train robots effectively, ensuring they can tackle multiple tasks with high success rates in real-life settings.
