□ The Department of Electrical Engineering and Computer Science at DGIST (President Kunwoo Lee) announced that Professor Kyung-joon Park’s research team has proposed the world’s first concept of Cyber-Physical AI (CPAI), a novel approach for effectively combining artificial intelligence with Cyber-Physical Systems (CPS). This research is expected to lay the foundation for AI to be reliably utilized in various real-world environments such as autonomous vehicles, smart factories, and medical robots.
□ NVIDIA CEO Jensen Huang recently emphasized that “the key to future technology lies in physical AI.” Physical AI refers to AI systems equipped with sensing and control components that directly operate in the physical world, including autonomous vehicles, robots, and smart factories. CPS refers to systems that tightly integrate physical devices with software and are widely used across domains such as smart factories, autonomous driving, and IoT-based infrastructure. However, many limitations have been raised regarding the integration of AI with CPS.
□ To ensure the reliable functionality of AI in real-world settings, several challenges must be addressed. First, AI malfunction can lead to severe consequences. For example, a self-driving car that fails to detect a pedestrian could cause a serious accident, and a medical robot error might pose a threat to a patient’s life. Second, inefficient AI performance within CPS may cause excessive computations, leading to faster battery depletion and delayed real-time decision-making. Lastly, the integration process of AI and CPS remains unsystematic, making it difficult to secure reliability.
□ In response, the DGIST research team has proposed Cyber-Physical AI (CPAI) as a systematic approach to facilitate the safe and efficient operation of AI within CPS. In defining CPAI, the team analyzed the concept across three dimensions: constraint, purpose, and approach, and systematically reorganized existing research. Moreover, they identified nine stages where problems could arise during AI implementation in CPS and proposed concrete solutions for each. In particular, the team addressed common challenges in AI-CPS integration, such as data bias, concept drift, and lack of trust, through experiments and case studies to seek effective resolutions.
□ This research marks the first attempt to comprehensively synthesize previously fragmented studies on CPS-AI integration, providing a structured framework that allows AI to operate reliably and resource-efficiently in physical environments. The CPAI concept is anticipated to significantly enhance the practical application of AI in various sectors, including autonomous driving, smart factory, and national defense technologies.
□ Professor Park stated, “For AI to operate as a trustworthy system in the real world, we must move beyond the mere application of technologies and instead focus on systematically designing the integration of AI and CPS.” He added, “This study organizes the previously scattered efforts and contributes to building safer and more efficient AI systems.”
□ This research was supported by funding from the Defense Acquisition Program Administration (DAPA) through the Korea Research Institute for Defense Technology Planning and Advancement (KRIT)’s specialized project on space surveillance and reconnaissance using heterogeneous satellite constellations, as well as the National Research Foundation of Korea’s Mid-Career Researcher Support Program. Professor Park, the corresponding author, also serves as the CTO of the robotics software startup S-Innovations Co., Ltd., and is actively engaged in the commercialization of CPAI.
- Corresponding Author E-mail Address : [email protected]