While traditional AI primarily analyses and generates information in the digital domain, Physical AI focuses on enabling systems to act purposefully in the real, physical world.
Physical AI represents a shift from rigid, preprogrammed robotic systems toward solutions that can adapt to new objects, situations, and tasks. Where traditional industrial automation often requires manual programming for each task, Physical AI makes it possible to develop generalizable strategies for manipulation, motion, and decision making. This is particularly relevant in environments where variation, uncertainty, or frequent changes make classical automation too inflexible. Much of this learning takes place in high fidelity simulation, where robots can explore thousands of variations in tasks, environments, and failures before being deployed in the real world.
How robots can use their entire body
Our research explores how robots can use their entire body—arm, base, and sensors—to plan and perform complex manipulation tasks, within what is known as embodied AI. A central focus is the development of learned manipulation policies that work across robots and tasks, as well as enabling robots to understand their surroundings and select appropriate actions based on context.
Identify and create opportunities for using Physical AI
SINTEF is a partner in the Norwegian Centre for Embodied AI and is responsible for two work packages: ‘Unifying Policies for Whole body Manipulation across Robot Embodiments’ and ‘Common sense Reasoning and Planning’. This provides a strong scientific foundation and a broad national and European network for developing next generation Physical AI technologies. For Norwegian industry, Physical AI opens new possibilities for flexible and adaptive automation. Through applied research and technology development, SINTEF helps identify and create opportunities for using Physical AI in production, logistics, aquaculture, and maintenance—showing how robots can complement people and contribute to increased safety, efficiency, and sustainability.
Embodied AI
“Embodied AI, refers to artificial systems whose cognitive processes emerge from continuous sensorimotor interactions in real-world environments” (Springer Nature, 2025) . “Physical AI lets autonomous systems like cameras, robots, and self-driving cars perceive, understand, reason, and perform or orchestrate complex actions in the physical world.” (Nvidia). As research in this area grows, terms like physical AI and embodied AI are in some cases used together, reflecting a shared ambition to build systems whose intelligence is grounded in real‑world interaction rather than remaining purely digital.