Date:
April 05, 2024
Covariant is focusing on developing an AI-based system that trains network robots to improve their picking capabilities as they learn. The system has proved invaluable in automating warehouses.
Covariant, a Bay Area firm that specializes in building AI that helps robots learn how to pick up different objects, has developed an AI-based system that trains network robots to improve their picking capabilities as they learn.
The system has proved invaluable in automating warehouses, which have to deal with an impossibly broad range of sizes, shapes, textures, and colors. Covariant's system has gone deeper into AI than any other robotic company by starting with pure software and pure AI, with the first hires being AI researchers.
The company takes a foundation model approach to learning, allowing it to train a large foundation-generalized model on more data to make the AI more generalized, unlike other AI models that are specific to a single use case.
Covariant is also working on building a foundation for future applications while focusing on picking and placing. The grasping capability or pick and place capability is definitely the first general capability that Covariant is giving robots.
