Dobb·E

About Dobb·E

Dobb·E is an open-source framework designed for home robotics, allowing users to teach robots household tasks efficiently. By utilizing a demonstration tool called the Stick, it collects data and trains models to adapt to new tasks quickly. Users benefit from a remarkable 81% success rate in varied environments.

Dobb·E offers free access to its software and models, with no subscription fees. This open-source platform empowers users to experiment with advanced robotics without financial barriers. The community-driven approach enhances collaboration, making Dobb·E a valuable resource for developers and researchers alike seeking home robotic solutions.

Dobb·E features a user-friendly interface designed for seamless navigation and efficient task management. Its clean layout, combined with intuitive functionalities, enables users to quickly start teaching robots new tasks. Dobb·E's responsive design simplifies interaction, fostering an engaging experience for both novices and experts in home robotics.

How Dobb·E works

Users interact with Dobb·E by first employing the demonstration tool, the Stick, to show their robots how to perform desired household tasks. After collecting about five minutes of data, the system processes this information to train a model swiftly. Within 20 minutes, users benefit from a policy that allows robots to tackle new tasks, achieving an impressive 81% success rate across varied environments.

Key Features for Dobb·E

Demonstration Collection Tool

Dobb·E's Demonstration Collection Tool, known as the Stick, revolutionizes how users teach robots. It enables quick data collection with affordable components, allowing for effective imitation learning. This innovative tool streamlines the process, enhancing the learning experience for robots in household environments.

Home Pretrained Representations (HPR)

Home Pretrained Representations (HPR) is a unique feature of Dobb·E. This pre-trained model allows robots to adapt quickly to new tasks using minimal data. By leveraging extensive training on diverse household tasks, HPR enhances the overall efficiency and effectiveness of robot performance in home settings.

Open-Source Accessibility

Dobb·E's open-source nature distinguishes it from other platforms, offering users free access to software, models, and hardware designs. This enables developers and researchers to collaborate, innovate, and contribute to advancing home robotics, fostering community growth and accessibility for all interested in robotic technologies.