The goal of the Adaptive Agentic AI research program is to push the boundaries of AI algorithms in the ability to adjust their actions based on new changes in the environment, thus creating more versatile AI agents that are able to make decisions autonomously.
Techniques we explore:
- Monte Carlo Tree Search (MCTS)
- Reinforcement Learning
- Transfer Learning
Published work:
- Kimiya Saadat, Richard Zhao. Enhancing Two-Player Performance Through Single-Player Knowledge Transfer: An Empirical Study on Atari 2600 Games. Proceedings of the Twentieth AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE-24), Lexington, USA, November, 2024. [Link] Acceptance Rate: 26.5%.
- Yiwei Zhang, Richard Zhao. Enhancing MCTS with Convolutional Autoencoder and Linear Approximator in XCOM-Inspired Environments. Proceedings of the Tenth AIIDE Workshop on Experimental Artificial Intelligence in Games (EXAG), Salt Lake City, Utah, USA, October, 2023. [Link] [BibTeX]
- Kimiya Saadat, Richard Zhao. Exploring Adaptive MCTS with TD Learning in miniXCOM. Proceedings of the Ninth AIIDE Workshop on Experimental Artificial Intelligence in Games (EXAG), Pomona, California, USA, October, 2022. [Link] [BibTeX]
- Hannah Ava Sloan, Richard Zhao, Faisal Aqlan, Hui Yang, Rui Zhu. Adaptive Virtual Assistant for Virtual Reality-based Remote Learning. American Society for Engineering Education Annual Conference & Exposition (ASEE 2022) Proceedings. Minneapolis, USA, June, 2022. [Link] [BibTeX]
