Real-time inference for robots at Physical Intelligence
Physical Intelligence (Pi) is developing a general-purpose robotic intelligence system using a Visual-Language-Action (VLA) model for real-time control. To enable rapid experimentation and the use of larger models, Pi utilizes Modal's cloud infrastructure for remote inference. They collaborated with Modal to create a specialized QUIC-based transport over UDP, minimizing network latency to approximately 10-15ms and avoiding the potential for TCP-related stalls in the robot's control loop. AI
IMPACT Enables more powerful AI models to be used for real-time robotic control by offloading computation to the cloud.