STATE16 RESEARCH

Physical AI Insights

Research, technical analysis and practical guidance on Physical AI assurance, VLA systems, world models, robotics safety and runtime guardrails.

A glowing optimization path bends through a dark scientific landscape, illustrating the direction-execution gap in physics-informed neural network optimization.

PHYSICAL AI

The Direction–Execution Gap in Physics-Informed Neural Network Optimization

Our IEEE Computational Intelligence Letters paper measures when a PINN optimizer step reverses its local descent signal - and how curvature exposes the gap.

Read article

Humanoid robot and robotic arm surrounded by layered Physical AI model representations

PHYSICAL AI

The Models Defining Physical AI

Explore the leading VLA, robot policy and world models defining Physical AI, including GR00T, pi0, Gemini Robotics, OpenVLA, Cosmos and V-JEPA.

Read article

Humanoid and industrial robot collaborating inside an integrated Physical AI production system

PHYSICAL AI

Physical AI Is Entering the Systems Phase

Physical AI is entering its systems phase. Dr. Barak Or examines robot architectures, data flywheels, safety, deployment and infrastructure.

Read article

Industrial humanoid robot operating inside a runtime safety envelope with a robotic arm and autonomous mobile robot

PHYSICAL AI

Physical AI Is Already Here. Where Are the Guardrails?

Physical AI is moving into real-world deployment. Dr. Barak Or explains why runtime guardrails, auditable decisions and assurance infrastructure are essential.

Read article