Reward AI introduces OM 1 control model for robotic hardware

Artificial intelligence startup Reward AI has unveiled OM 1, a foundational control model designed to help robots interact with the physical world more like humans do. Instead of writing custom software for every specific piece of hardware, the company developed a unified system that can operate across different physical designs, ranging from standard industrial arms to two-legged humanoid machines.

To train the system, Reward AI created a specialized wearable glove that captures detailed human movements during everyday tasks like cooking or sorting items. The wearable records hand positions, applied force, tactile feedback, and precise motion tracking without requiring demonstrators to slow down. By converting fluid human motion into standardized training data, the model attempts to give different types of machinery the speed, dexterity, and delicate touch needed to work alongside people.

While training machines with human movement data is a promising approach, moving from controlled tests to chaotic everyday environments remains a major challenge. It is still unclear how effectively this unified model handles hardware wear over time or unexpected objects outside its training data without requiring manual adjustments. If general control software like this succeeds, it could significantly lower the barrier to deploying multi-purpose robots in warehouses and factories.