3002677619

Controlling multiple, heterogeneous agents engaged in cooperative goals is an important problem with many applications, e.g. target search, scene identification, object tracking, and quantity estimation problems. The computational challenges for this class of problems are numerous and substantial, with difficulties arising separately and conjointly in inference, learning, decision-making, and communication. Despite significant advances in recent years in solving these problems, state-of-the-art algorithms still far lag behind human performance in many respects, in particular in terms of context-sensitivity, robustness, goal-directness, and dynamic adaptability. Based on my group's existing experience and interest in cognitive and computational neuroscience, we anticipate that our expertise will complement other team members' by contributing to the project with the three following objectives: (1) understanding the computational processes underlying human cognition, (2) developing computational frameworks and algorithms for solving canonical problems in multi-agent decision-making, (3) identifying opportunities for productive interactions between humans and artificial agents. Year 4 Goals: model-data comparison

San Diego University Of California

$40,000.00

May 26, 2015

La Jolla, California 92093-0934 United States

Regents Of The University Of Michigan

W911NF1110391

VALUE-CENTERED INFORMATION THEORY FOR ADAPTIVE LEARNING, INFERENCE, TRACKING AND EXPLOITATION

12.431 - Basic Scientific Research

Department of Defense (DoD)

Project Grant

Ann Arbor, Michigan 481091274 United States