Agility Robotics has announced an expanded collaboration with NVIDIA to further integrate NVIDIA’s software and hardware into the training and simulation of its humanoid robot, Digit. As part of the partnership, Agility Robotics is increasing its adoption of NVIDIA Isaac Sim and NVIDIA Isaac Lab, simulation and learning frameworks designed to develop and test Digit’s behaviors in virtual environments. The company is also working with NVIDIA to make Digit models available to partners through Mega, an NVIDIA Omniverse Blueprint.
The integration of NVIDIA’s technology allows Agility Robotics to enhance Digit’s capabilities, including step-recovery improvements that help the robot maintain stability in response to physical disturbances. The collaboration also enables Digit to utilize NVIDIA’s AI acceleration platform for real-time perception and decision-making, essential for interactions in dynamic environments.
Agility Robotics’ Chief Technology Officer, Pras Velagapudi, emphasized the benefits of the collaboration, stating that NVIDIA’s focus on physical AI is accelerating the development of Digit’s capabilities. Rev Lebaredian, Vice President of Omniverse and Simulation Technology at NVIDIA, highlighted the role of simulation and reinforcement learning in advancing humanoid robotics and enhancing Digit’s ability to understand and interact with real-world environments.
In addition to the expanded use of NVIDIA’s simulation tools, Agility Robotics has received early access to NVIDIA Cosmos, a platform for generative world foundation models, and Mega, which facilitates large-scale testing and optimization of physical AI in digital twins. Agility has already begun using these technologies in collaboration with automotive manufacturing customer Schaeffler to digitally replicate facility assets such as bins and shelves, streamlining robot deployment and testing.
Digit, described by Agility Robotics as the most advanced Mobile Manipulation Robot (MMR) on the market, is actively deployed in warehouses and factories, addressing automation challenges in existing facilities. The robot is designed to navigate human-centered environments and handle workflows that have traditionally been difficult to automate.
