Image

Shlomo Zilberstein, professor in the Manning College of Information and Computer Sciences (CICS), has received the 2025 ACM/SIGAI Autonomous Agents Research Award for his significant and influential work in the field of autonomous agents. The award will be presented at the 24th International Conference on Autonomous Agents and Multiagent Systems, May 19-23 in Detroit.

Zilberstein is cited as having “established the field of decentralized Markov Decision Processes (DEC-MDPs), laying the groundwork for decision-theoretic planning in multi-agent systems and multi-agent reinforcement learning (MARL).”

Zilberstein’s research centers on automated planning, enabling autonomous systems, such as service robots and self-driving vehicles, to intelligently assess their environments and make informed decisions. His work focuses on long-term autonomy, enabling AI systems to maintain reliable performance over extended periods by adapting to changing environments, managing resources efficiently and requiring minimal human oversight.

The DEC-MDP framework was first introduced in the 2000 paper, “The Complexity of Decentralized Control of Markov Decision Processes,” presented at the Conference on Uncertainty in Artificial Intelligence and has since received more than 2000 citations and earned the IFAAMAS Influential Paper Award in 2019. Stemming from the synthesis project by Daniel Bernstein, then Zilberstein’s doctoral advisee, the seminal paper provided the first rigorous mathematical framework for sequential decision-making in scenarios involving multiple decision-makers with partial information.

The research was also the first to characterize the foundational theoretical differences between centralized and decentralized MDPs, establishing computational complexity benchmarks, and developing the first exact and approximate dynamic programming algorithms for these problems.

Zilberstein’s collaborations with industry have helped translate his research into impactful, real-world applications. Notably, his MODIA (Multiple Online Decision-components with Interacting Actions) system has been deployed in Nissan’s experimental autonomous vehicles in the U.S. and Japan, enabling efficient interactions with numerous dynamic entities such as vehicles and pedestrians. To date, more than 20 patents based on this work have been filed, with a dozen already granted.

“I am deeply honored to receive this recognition from the ACM/SIGAI community,” said Zilberstein. “The journey of developing foundational models for multi-agent planning has been incredibly rewarding, and I am grateful for the many students, collaborators and colleagues who contributed to this work. It is exciting to see how ideas that began with theoretical exploration continue to influence the future of autonomous systems.” 

Zilberstein joined the CICS faculty in 1993. He is a fellow of the Association for the Advancement of Artificial Intelligence (AAAI) and the Association for Computing Machinery (ACM). The recipient of the UMass Amherst Chancellor’s Medal and the AAAI Distinguished Service Award, he is currently a trustee of the International Joint Conference on Artificial Intelligence (IJCAI), chair of IJCAI-25 and the chairman of the AI Access Foundation.