A Systematic Review of Physical Artificial Intelligence (Physical AI): Concepts, Applications, Challenges, and Future Directions
Renny Sari Dewi, Afif Nurul Kawakib, Maghfiroh Nur Laili, Anannda Lailatul Fauziah, Salsabilla Rahma Sabrina, Rafidah Latifah Hana · Journal of Artificial Intelligence and Engineering Applications (JAIEA) · 2025
Physical AI represents a significant evolution from digital AI , interacting directly with the physical world and mimicking human functions to a greater extent. PAI is a multidisciplinary field divided into Integrated PAI (IPAI) and Distributed Physical AI (DPAI). This systematic literature review analyzes the concept of PAIs, their implementation in various domains such as IoT, automotive, agriculture, healthcare, and logistics, and highlights their transformative potential. Nonetheless, PAIs face significant challenges such as general AI concerns (privacy, bias) and specific challenges (presence in unregulated spaces, information organization, social acceptance, Cannikin law). The integration of PAIs into Cyber-Physical Systems (CPS) also presents challenges related to uncertainty, limited resources, and adversarial attacks. PAIs are supported by advanced technologies from materials science, mechanical engineering, computer science, chemistry, and biology, including deep learning, multimodal processing, domain randomization, zero-shot learning, and large language objects (LLOs). This research provides comprehensive insights to drive the development of reliable and transformative PAIs in the future.