THE ROLE OF PHYSICS ENGINES IN CGI: AI-POWERED REALISM IN ANIMATION

Krunal Suthar; Yogesh Patel; Harshad Chaudhary; Mitul Patel; Shraddha Modi

doi:10.29121/shodhkosh.v5.i6.2024.4782

Authors

Krunal Suthar Assistant Professor, Government Engineering college, Patan, Gujarat, India
Yogesh Patel Assistant Professor, Government Engineering college, Patan, Gujarat, India
Harshad Chaudhary Assistant Professor, Government Engineering college, Patan, Gujarat, India
Mitul Patel Assistant Professor, Government Engineering college, Patan, Gujarat, India
Shraddha Modi Assistant Professor, L. D. college of Engineering, Ahmedabad, Gujarat, India

DOI:

https://doi.org/10.29121/shodhkosh.v5.i6.2024.4782

Keywords:

Artificial Intelligence, Computer-Generated Imagery, Physics, Animation, Deep Learning

Abstract [English]

The integration of physics engines and artificial intelligence (AI) has revolutionized the field of computer-generated imagery (CGI), enabling the creation of highly realistic animations. Traditional physics engines simulate real-world behaviors such as gravity, fluid dynamics, and collision detection, but they often require extensive manual fine-tuning by animators to achieve the desired realism. AI-powered techniques, particularly deep learning models and reinforcement learning, have emerged as powerful tools to optimize and enhance these simulations, leading to improved efficiency and realism. This paper explores the role of physics engines in CGI, the challenges associated with traditional approaches, and the advancements enabled by AI. We analyze existing techniques and propose an AI-driven framework that leverages deep reinforcement learning to optimize physics simulations for CGI applications. This framework aims to reduce manual intervention while achieving a more lifelike representation of real-world physics, thereby streamlining the animation production pipeline and enhancing the overall visual fidelity of animated films. Through an extensive review of literature and analysis of various AI-enhanced physics engines, we provide a comprehensive understanding of the interplay between physics, AI, and CGI.

References

S. Patil, "Advancements in AI-driven CGI physics simulations," IEEE Transactions on Visualization and Computer Graphics, vol. 27, no. 8, pp. 1324-1340, 2023.

D. Baraff and A. Witkin, "Numerical methods for rigid body dynamics," ACM Transactions on Graphics, vol. 18, no. 3, pp. 245-262, 1998.

M. Müller, B. Heidelberger, M. Teschner, and M. Gross, "Position based dynamics," Journal of Computer Animation and Virtual Worlds, vol. 15, no. 1, pp. 69-77, 2004. DOI: https://doi.org/10.1002/cav.18

O. Hilliges et al., "Fluid simulation techniques for CGI," ACM SIGGRAPH Conference Proceedings, pp. 112-120, 2017.

X. Wu et al., "Deep learning for physics-based animation," IEEE Transactions on Neural Networks and Learning Systems, vol. 31, no. 12, pp. 2302-2315, 2022.

J. Holden et al., "Deep reinforcement learning for character animation," IEEE Transactions on Image Processing, vol. 30, no. 5, pp. 845-860, 2021.

R. S. F. Chan, T. Kim, and M. L. L. Zhou, "Interactive fluid simulation for computer graphics," ACM Transactions on Graphics, vol. 28, no. 4, pp. 1-7, 2009.

J. C. McCrae and P. D. Hooker, "Simulating rigid body dynamics for real-time applications," ACM SIGGRAPH Asia Conference Proceedings, pp. 32-45, 2015.

M. A. D. Valente and R. S. F. Chan, "Real-time water simulations using deep learning," IEEE Transactions on Visualization and Computer Graphics, vol. 24, no. 5, pp. 1656-1669, 2018.

C. E. J. M. Oliveira, "Physics-driven character animation using machine learning," ACM Transactions on Graphics, vol. 37, no. 1, pp. 1-11, 2018.

R. S. Jung, C. Kim, and S. K. Hwang, "Reinforcement learning in 3D rigid body simulations," IEEE Transactions on Neural Networks and Learning Systems, vol. 30, no. 3, pp. 592-605, 2019.

C. Yang, Y. Zhang, and D. S. L. Man, "Physics-based animation with deep learning models," IEEE Transactions on Image Processing, vol. 28, no. 8, pp. 1453-1465, 2019.

M. S. Abolhasani and K. M. M. Kalantari, "AI-driven fluid dynamics simulation for realistic CGI," ACM SIGGRAPH Conference Proceedings, pp. 203-210, 2020.

A. K. Sharma, M. A. Barros, and J. G. Martinez, "Machine learning for collision detection in virtual environments," ACM Transactions on Graphics, vol. 39, no. 5, pp. 1-9, 2020.

A. G. Saldaña et al., "Real-time character simulation and reinforcement learning for CGI," IEEE Transactions on Computational Intelligence and AI in Games, vol. 12, no. 2, pp. 34-45, 2020.

L. Wang and X. Xu, "AI-based optimization of soft body dynamics in CGI applications," IEEE Transactions on Visualization and Computer Graphics, vol. 26, no. 7, pp. 2095-2105, 2020. DOI: https://doi.org/10.1109/TVCG.2019.2935603

M. S. Ribeiro and M. T. N. Almeida, "Simulating materials in CGI using deep reinforcement learning," ACM Transactions on Graphics, vol. 41, no. 4, pp. 1-13, 2022.

J. Chen and A. Banerjee, "Advances in AI-powered animation physics for gaming and film production," Journal of Computer Graphics Techniques, vol. 11, no. 2, pp. 144-157, 2022.

K. L. Zhou et al., "Physics-based deep learning models for animation in digital environments," IEEE Transactions on Neural Networks and Learning Systems, vol. 33, no. 6, pp. 1990-2003, 2022. DOI: https://doi.org/10.1109/TNNLS.2021.3059529

S. Arif and R. H. Reiter, "Integrating deep learning with fluid simulation techniques for interactive CGI," IEEE Transactions on Image Processing, vol. 31, no. 8, pp. 1715-1727, 2023.