Hello world, meet Anna – our first prototype of an affordable, general-purpose household robot 🤖
It uses a combination of LLMs, VLMs, and VLAs to understand language, perceive surroundings, and automate 🏠 chores seamlessly.
We chose the wheeled form factor over bipedal for its simple yet functional design. For the first generation embodied AI in homes, we believe this form factor wil be friendly and less intimidating than the alternatives.
A simpler design also helps with our goal of making these robots affordable. We are hoping to ship these devices for under $5k
In the coming weeks, we will share an updated prototype that will be closer to our production design and includes vertical movement that is critical for most household chores.
⬇️ is a quick demo of Anna, responding to voice and handing tasks end to end
Hello world, meet Anna – our first prototype of an affordable, general-purpose household robot 🤖
It uses a combination of LLMs, VLMs, and VLAs to understand language, perceive surroundings, and automate 🏠 chores seamlessly.
We chose the wheeled form factor over bipedal for its simple yet functional design. For the first generation embodied AI in homes, we believe this form factor wil be friendly and less intimidating than the alternatives.
A simpler design also helps with our goal of making these robots affordable. We are hoping to ship these devices for under $5k
In the coming weeks, we will share an updated prototype that will be closer to our production design and includes vertical movement that is critical for most household chores.
⬇️ is a quick demo of Anna, responding to voice and handing tasks end to end
Robots that generalize to the real world need lots of diverse training data to build robust models. But collecting that data is a huge challenge.
That’s why we built TeleOp-XR: an open system to teleoperate robots in Augmented Reality! 🤖✨
A quick demo and instructions 🧵⬇️
Ready to collect better data for your robots?
- Clone the repo from https://t.co/cY0AoZ47tQ
- Add your robot config
- Start teleoperating and recording in AR
Towards generalizable robots—one episode at a time! 🚀
Robots that generalize to the real world need lots of diverse training data to build robust models. But collecting that data is a huge challenge.
That’s why we built TeleOp-XR: an open system to teleoperate robots in Augmented Reality! 🤖✨
A quick demo and instructions 🧵⬇️
TeleOp-XR is modular:
- Support new robots by editing a config file
- Define custom AR menus for your tasks and bind them to actions in the backend
- Plugin-like architecture to easily support new robot backends
@pthangeda_@chris_j_paxton We built something similar for data collection at Vexa.. plan to clean it up a bit and open source it later this week... It should work out of the box with @LeRobotHF and can be extended for others
@hubermanlab@NIH@CDCgov Is it possible let the public directly fund the research? Maybe a crowd source platform similar to Kickstarter and then alllow the donors claim some tax benefits.. that way we can hopefully make the system more efficient
Just started writing an MCP server for FreeCAD and thought I will do a quick search before going too far
Aaand yes, looks like someone just wrote one
https://t.co/jQnu1SNDma
NVIDIA announced GR00T N1, the first fully customizable open-source humanoid robot foundation model, designed to advance general-purpose robotics.
⦿ A dual-system AI inspired by human cognition: System-1 handles fast, intuitive actions, while System-2 enables methodical decision-making. System-1 is trained on human demonstration data and a massive amount of synthetic data.
⦿ GR00T N1 can easily generalize across common tasks - such as grasping, moving objects with one or both arms, and transferring items from one arm to another - or perform multistep tasks that require long context and combinations of general skills.
⦿ CEO of 1X: "With a minimal amount of post-training data, we were able to fully deploy on NEO Gamma - furthering our mission of creating robots that are not tools, but companions that can assist humans in meaningful, immeasurable ways."
⦿ Early adopters also include Agility Robotics, Boston Dynamics, Mentee Robotics, and NEURA Robotics.