Physical Intelligence on the Edge: A Vision for the Decade Ahead

This article examines key challenges in computing systems research under the emerging paradigm of Physical Intelligence on the Edge (PIE), in which raw sensor streams are transformed into real-time, safety-critical intelligence that can act in the physical world. It traces the evolution of computing architectures from centralized systems to distributed systems and edge computing, and argues that PIE constitutes a qualitative shift: the edge becomes the primary platform for tightly integrating sensing, reasoning, and actuation under stringent real-time constraints. The article identifies five emerging research thrusts: embodied spatial reasoning, embodied tem- poral reasoning, edge-native customization, symbiosis, and sustainability. Using a hypothetical PIE scenario, it highlights a critical gap between current systems and future PIE-enabled auton- omy: today’s edge platforms can run individual components of perception and inference, but they still struggle to autonomously close the sense–think–act loop with certifiable timing and safety guarantees. The article concludes by arguing for a shift in systems thinking from efficiently mov- ing and processing data (bits) to predictably influencing the physical world (atoms), positioning edge-native system design as a foundation for next-generation autonomous and robotic systems.