What is it about?
Imagine you have real-world machines and devices (like robots, cars, or factory equipment) all connected to computer systems that can sense and react to what’s going on around them. These setups are called Cyber-Physical Systems (CPS). When we add Artificial Intelligence (AI) into the mix, we open the door to amazing possibilities—like smarter automation and greater efficiency—but we also run into big questions about safety and resources. Here’s why: Dealing with Uncertainty AI focuses on increasing the chances of good outcomes, but it doesn’t guarantee absolute safety. In contrast, CPS must make sure that “all bad things must not happen,” which places a strong emphasis on avoiding any dangerous situations. This difference means we have to carefully manage uncertainty whenever AI is integrated into real-world systems. Managing Limited Resources Many CPS devices, such as sensors or small robots, have limited power, memory, or processing ability. AI often works best when it has plenty of data and computing resources. Bringing AI into these constrained environments requires us to be very aware of resource usage to ensure the system remains efficient and reliable. That’s where Cyber-Physical AI (CPAI) comes in. CPAI is a new research field that combines AI with CPS. We organize CPAI using three main ideas—Constraint (C), Purpose (P), and Approach (A)—and we break down the CPS-AI integration process into 3 phases and 9 steps to keep things systematic. After reviewing 104 different research papers, we discovered 9 major challenges and insights that highlight the future direction of CPAI. The goal of CPAI is simple: help people reliably and efficiently use AI as part of real-world machines and systems. By bringing different studies together under one roof, CPAI aims to make these technologies safer, smarter, and more practical for everyday use.
Featured Image
Photo by Igor Omilaev on Unsplash
Why is it important?
Our work stands out because it addresses the immediate need to merge AI’s powerful decision-making with the strict safety and resource requirements of real-world systems. This is especially relevant given current industry trends, such as NVIDIA CEO Jensen Huang’s vision of “Physical AI,” which focuses on bringing AI into tangible products and environments—like autonomous machines, advanced robotics, or smart factories. To make this vision a reality, we need a systematic understanding of how AI interacts with the physical world, from managing energy constraints to ensuring no catastrophic failures occur. CPAI fills this gap by offering a structured way to reliably and efficiently embed AI into CPS, ultimately guiding researchers and practitioners to create safer, more effective physical machines that harness the full potential of AI.
Perspectives
While fundamental AI research—covering algorithms, models, and theory—continues to evolve at a breathtaking pace, I’ve noticed that relatively few in academia focus on a different but equally critical question: How do we effectively integrate AI into existing real-world systems? In my view, this “integration research” is just as important as foundational AI work. Without viable pathways to embed advanced AI components into practical settings—such as vehicles, industrial equipment, or healthcare devices—the potential benefits of cutting-edge algorithms risk being delayed or underutilized. This publication reflects my conviction, as well as that of my colleagues in the Cyber-Physical Systems community: if we don’t tackle the integration challenge head-on, we won’t be able to fully unlock AI’s transformative power. Through the concept of Cyber-Physical AI (CPAI), we provide a structured framework that we hope will accelerate efforts to bring AI into physical reality. My personal hope is that by defining CPAI as its own domain, more researchers and practitioners will be encouraged to explore how advanced AI can work reliably, safely, and efficiently alongside the hardware and resource constraints of real-world environments. Ultimately, I believe this convergence of AI and CPS will open the door to a new era of intelligent, connected systems that profoundly benefit our daily lives.
sanghoon lee
Daegu Gyeongbuk Institute of Science and Technology
Read the Original
This page is a summary of: Cyber-Physical AI: Systematic Research Domain for Integrating AI and Cyber-Physical Systems, ACM Transactions on Cyber-Physical Systems, April 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3721437.
You can read the full text:
Contributors
The following have contributed to this page
