Professor Donald Perlis receives $1M DARPA Award for leading AI Research

Descriptive image for Professor Donald Perlis receives $1M DARPA Award for leading AI Research

Professor Donald Perlis along with his team, Justin Brody and Timothy Clausner (both visiting scientist, UMIACS) received $980 K Defense Advanced Research Projects Agency (DARPA) award for their project titled -- "Toward Knowledge of Cooperative Agency: A Foundation for Task-General Teaming.”

The project will take the first steps towards an initial theory of cooperative agency as the bedrock on which general autonomous teaming can be built.

The DARPA award came through the recently launched CREATE (Context Reasoning for Autonomous Teaming) program, which aims to explore the utility of Artificial Intelligence (AI) on the autonomous formation of scalable machine-to-machine teams.
CREATE program focuses to develop Artificial Intelligence (AI) technologies in support of decentralized autonomous team formation enabling heterogeneous context-aware agents to satisfy multiple simultaneous and unplanned mission goals.

“While the capacity of agents to interact meaningfully has advanced significantly, most such work focuses either on action, knowledge, and inference specific to particular task-types, or on theoretical idealizations that are far removed from the real-time needs of real-world agents” explained Perlis.

The awarded project proposes to develop KoCA (Knowledge of Cooperative Agency) technology to build a solid theoretical foundation for real-time context-aware cooperating agents. An agent that is part of a team will need not only to have access to its own beliefs and intentions and those of others, it will also need to understand how such beliefs and intentions evolve and interact over time and during actions.

Professor Darsana Josyula from the Bowie State University serves as a Co PI on this project.

In addition to being a full professor of computer science, Perlis holds joint appointments in UMIACS and NACS.

The Department welcomes comments, suggestions and corrections.  Send email to editor [-at-] cs [dot] umd [dot] edu.