Developing Responsive And Adaptive Artificial Agents for Team Training

Teams form the fundamental structure of military forces and effective team training is therefore critical for mission success. Such training requires that training scenarios include intact teams and promote self-guided learning via simulated mission contexts. A major challenge to engaging in such training is ensuring that sufficient numbers of active duty personnel are able to participate, and team training programs are tailored to the individual needs of trainees. One way to overcome these challenges is to incorporate artificial agents (AAs) within the training context.

The effectiveness of human-AA team training depends on the ability of AAs to adapt to human co-actors in a seamless manner. In order to enhance real-world outcomes, AAs must also incorporate natural, human-like patterns of behavioural action, decision-making, and communication. As such, ensuring effective human-AA team training requires modelling the behavioural dynamics of successful human performance and then implementing these models within the control architecture of AAs.

According, the proposed project has two primary research AIMS:

AIM 1: Demonstrate how human performance and communication within complex team settings can be modelled using a hierarchical framework of dynamical action and decision-making primitives and generative deep-learning based NLP methods.

AIM 2: Demonstrate how hierarchical models of complex human performance and communication (composed of dynamical action and decision-making primitives, and deep-learning based NLP methods) can be employed to develop AAs capable of effective team training within tactical action and command-and-control task contexts.

RECENT PUBLICATIONS:

Rigoli, L. M., Nalepka, P., Douglas, H., Kallen, R. W., Hosking, S., Best, C., & Richardson, M. J. (2020, May). Employing Models of Human Social Motor Behavior for Artificial Agent Trainers. In Proceedings of the 19th International Conference on Autonomous Agents and MultiAgent Systems (pp. 1134-1142).

Nalepka, P., Kallen, R. W.  Chemero, A., Saltzman, E., & Richardson, M. J., (2019). Practical Applications of Multiagent Shepherding for Human-Machine Interaction. Paper for the 17th International Conference on Practical Applications of Agents and Multi-Agent Systems (PAAMS).