A Study on Recent Research Trends and Future Directions in Physical AI (Physical AI의 최근 연구 동향과 발전 방향에 대한 연구)

JongHoon Kim (김종훈); Eui-Jik Kim (김의직); Dongwan Kim (김동완)

doi:10.20465/KIOTS.2025.11.4.107

A Study on Recent Research Trends and Future Directions in Physical AI

Journal of Internet of Things and Convergence
Abbr : JKIOTS
2025, 11(4), pp.107~116
DOI : 10.20465/KIOTS.2025.11.4.107
Publisher : The Korea Internet of Things Society
Research Area : Engineering > Computer Science > Internet Information Processing
Received : July 18, 2025
Accepted : August 15, 2025
Published : August 31, 2025

JongHoon Kim ¹, Eui-Jik Kim ², Dongwan Kim ³

¹동아대학교
²한림대학교
³동아대학교 전자공학과

Accredited

ABSTRACT

Physical AI refers to technologies enabling mechanical systems—such as robots and drones—to learn their operational dynamics and embed intelligence. This field has attracted significant attention in autonomous robotic systems engineering, where machines learn and execute complex tasks in real-world environments. In this paper, we compare QT-Opt, Dreamer, Gato, and RoboCat across algorithm, data requirements, computational complexity, Sim-to-Real, and limitations. We also review domain randomization and world-model–based self-supervised learning (SGF), and a surgical-robot case. Furthermore, we assess V-JEPA 2 and NVIDIA Cosmos in terms of data scale, use cases, and constraints. Finally, we discuss future development strategies based on the limitations identified in recent studies.

KEYWORDS

Artificial Intelligence(AI), Physical AI, Autonomous Robotics Systems, AI Convergence, Sim-to-Real

Citation status

* References for papers published after 2025 are currently being built.

[confproc] I.Kaur / 2023 / Survey on Computer Vision Techniques for Internet-of-Things Devices / 2023 IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology(IAICT) : 244~250

[journal] J.Arents / 2022 / Smart Industrial Robot Control Trends, Challenges and Opportunities within Manufacturing / Applied Sciences 12(2) : 937~

[journal] A.Miriyev / 2020 / Skill for Physical Artificial Intelligence / Nature Machine Intelligence 2(11) : 658~660

[other] F.Bousetouane / 2025 / Physical AI Agents: Integrating Cogni-tive Intelligence with Real-World Action : 1~27

[confproc] D.Kalashnikov / 2018 / QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation / Proceedings of The 2nd Conference on Robot Learning (COLR) 87 : 651~673

[confproc] D.Hafner / 2020 / Dream to Control: Learning Behaviors by Latent Imagination / Proceedings of the International Conference on Learning Representations(ICLR) : 1~20

[journal] S.Reed / 2022 / A Generalist Agent / Transactions on Machine Learning Research(TMLR) (371) : 2835~8856

[journal] K.Bousmalis / 2024 / RoboCat: A Self-Improving Generalist Agent for Robotic Manipulation / Transactions on Machine Learning Research(TMLR) (1540) : 2845~8856

[other] A.A.Malik / 2023 / Intelligent Humanoids in Manufacturing to Address Worker Shortage and Skill Gaps: Case of Tesla Optimus : 1~20

[confproc] J.Tobin / 2017 / Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World / Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems(IROS) : 23~30

[confproc] J.Robine / 2025 / Simple, Good, Fast: Self-Supervised World Models Free of Baggage / Proceedings of the International Conference on Lear-ning Representation(ICLR) (10365) : 1~21

[confproc] A.Bardes / 2022 / VICReg: VarianceInvariance-Convariance Regularization for Self-Supervised Learning / Proceedings of the International Conference on Learning Representations(ICLR) : 1~22

[journal] Y. Ou / 2023 / Sim-to-Real Surgical Robot Learning and Autonomous Planning for Internal Tissue Points Manipulation Using Reinforcement Learning / IEEE Robotics and Automation Letters 8(5) : 2502~2509

[book] F.Faure / 2012 / SOFA: A Multi-Model Framework for Interactive Physical Simulation / Springer : 283~321

[other] M.Assran / 2025 / V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning : 1~48

[other] NVIDIA / 2025 / Cosmos World Foundation Model Platform for Physical AI : 1~75

KJCKorea
Journal Central

Journal of Internet of Things and Convergence 2025 KCI Impact Factor : 0.75

A Study on Recent Research Trends and Future Directions in Physical AI

ABSTRACT

KEYWORDS

Citation status

* References for papers published after 2025 are currently being built.

Journal of Internet of Things and Convergence 2025 KCI Impact Factor : 0.75

A Study on Recent Research Trends and Future Directions in Physical AI

ABSTRACT

KEYWORDS

Statistics

Tools

Issue List

Citation status

KCI Citation Counts (0)

REFERENCES (16) * References for papers published after 2025 are currently being built.

Search PDF

Citation

* References for papers published after 2025 are currently being built.