A Physical AI system gathers sensor data about its environment, builds a representation of it, plans an action (e.g. a motion trajectory) and executes it through actuators, then corrects behavior based on feedback. Its core is AI models (perception, world models, control policies) embedded in a physical platform with a control loop.
Classical AI operates mainly on data in the digital domain; Physical AI addresses the challenge of acting in the unpredictable, continuous physical world, where perception, planning and execution must be coupled in a real-time loop.
