Dobb·E

About Dobb·E

Dobb·E offers an innovative open-source framework that empowers robots to learn household tasks quickly and efficiently. Targeted towards researchers and developers, Dobb·E uses a unique demonstration collection tool, the Stick, for adaptation and training, providing a seamless way to enhance robotic home assistance.

Dobb·E offers free access to its open-source framework, with no subscription tiers. Users benefit from community support and the ability to customize the system for varied household tasks. This flexibility encourages collaborative upgrades and developments in home robotics, enhancing overall functionality.

Dobb·E features a user-friendly interface designed for simplicity and efficiency. The layout ensures intuitive navigation, making it easy for users to access its core functionalities and tools. The seamless experience fosters engagement and effectiveness, essential for effective robot training and task accomplishment.

How Dobb·E works

Users interact with Dobb·E by first utilizing the Stick to gather demonstration data over five minutes. This data trains the Home Pretrained Representations model to adapt to new tasks. Then, within fifteen minutes, users can implement the policy generated by Dobb·E to achieve an average success rate of 81%, simplifying household robotics.

Key Features for Dobb·E

Demonstration Collection Tool

The Stick is a revolutionary demonstration collection tool integral to Dobb·E. It allows users to easily collect task demonstrations at home using affordable components. This unique approach significantly enhances data collection efficiency, paving the way for effective task learning and adaptation in household robotics.

Home Pretrained Representations (HPR)

Home Pretrained Representations (HPR) is foundational to Dobb·E, serving as a pre-trained model that accelerates robot training. Utilizing a self-supervised learning approach, HPR enables efficient task adaptation in diverse environments, thereby enhancing the overall performance and versatility of robotic systems in homes.

Open-source Accessibility

Dobb·E is fully open-source, promoting innovation and accessibility in home robotics. By providing its software, hardware designs, and datasets for public use, Dobb·E enables a collaborative community to advance robotic learning and practical applications, making household tasks manageable across different environments.