Developing a Robust Architecture for Cooperative Tasks
Abstract
The goal of this project is to extend the ACT-R architecture, and its learning capabilities so that it can provide an understandingof collaboration in tasks that involve multiple agents. This will involve understanding how one agent represents the behavior of other agents and the dependencies between its behavior and others. We will investigate the role of communication in this process and how one agents learning can be facilitated by other agents. We will extend current work on tracking cognition with EEG to relatethe activity of multiple agents, exploring co-information that combines the signals from multiple agents.We will work with a variant of the Space Fortress game called Coop Space Fortress which involve multiple agents collaborating to destroy fortresses. We start with successful ACT-R models of learning that we have developed for the single agent version of Space fortress. We will need extend those models with representations of other agents and interdependency ofactions. We will add limited communication possibilities to the game and limited language processing capabilities to our model to understand the contribution of communication. We will extend our Sketch-and-Stitch method for EEG to track the cognition of multiple agents, wearing practical EEG headsets. In addition,we will explore whether success of collaboration is related to measures of brain synchrony and other measures of co-information in the joint signal.The goal of this research is to provide a deeper understanding of collaborative performance and learning. As suchit provides a basis for understanding how to maximize team performance and to design environments and instruction to optimize learning. The computational nature of the models facilitates understanding how to integrate human and synthetic teammates. If successful, this project will enable computational analysis to design optimal team composition and optimal training for collaborative teams which are essential to many Naval operations. Understanding how the neural signals relate to learning and performance will provide the basis for using practical EEG signals to enhance collaboration among human teams and also teams of human and artificial agents. "Approved for public release"
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 05, 2021
- Source ID
- N000142112586
Entities
People
- John R. Anderson
Organizations
- Carnegie Mellon University
- Office of Naval Research
- United States Navy