Physical Intelligence (π) introduces π0, a general-purpose robot foundation model designed to bridge the gap between AI and the physical world. Key features include:
- Versatile Robot Control: Enables robots to perform a wide range of tasks through text instructions, similar to LLMs.
- Multimodal Training: Trained on images, text, and actions, acquiring physical intelligence through embodied experience.
- Cross-Embodiment Learning: Controls various robot types by learning low-level motor commands via a novel architecture.
- Internet-Scale Semantic Understanding: Inherits semantic knowledge from pre-trained vision-language models (VLMs).
- Flow Matching for Dexterity: Augments VLMs with continuous action outputs for high-frequency motor control.
Use cases include automating complex tasks like laundry folding, table bussing, and box assembly, demonstrating adaptability and problem-solving in unstructured environments.
