The Blanket
Robots that think before they act.
Grounding robot foundation models in probabilistic world
models
to enable adaptive, robust, and safe robots.
auto·motion
From product design to robot assembly, without reprogramming.
Load Assembly Design
Operator uploads design file (e.g., CAD file) into the system
Plan & Execute
auto·motion generates a hierarchical task execution model and starts building
Query when needed
When uncertain, the system queries an operator for targeted input
human-in-the-loopLearn & Adapt
The system adapts online, recovers from errors, and improves with use.
Discussing with manufacturers, two major cost drivers stand out: 1) frequent reprogramming, resulting in downtime and often requiring external specialists, and 2) the need to engineer environments around robot limitations. Instead of adapting environments to robots, auto·motion makes robots adapt to real-world environments.
Technology
We are building a new intelligence framework that enables robots to operate reliably in complex physical environments without needing constant reprogramming. Our system architecture grounds robot foundation models in continuously learning probabilistic world models using active inference, allowing robots to reason and plan under uncertainty, adapt in real time, recover from failure, and keep producing with minimal supervision.
Contact
Reach out to us at contact@theblanket.ai if you would like to invest or are interested in our technology.