Learning from Demonstration in a Military Drill: Artificial Intelligence of Maneuver and Mobility – Human Agent Teaming Integrated Capability Experiment (AIMM-HAT ICE)

Author(s):

Angelique Scharine, David Chhan, Brian Cesar-Tondreau, Vernon Lawhern, Joe Rexwinkle, and Julia L. Wright

Abstract:

The Artificial Intelligence of Maneuver and Mobility – Human Agent Teaming Integrated Capability Experiment (AIMM-HAT ICE) explored novel autonomous capabilities through a three-part experiment. AI navigation models typically require large datasets for training and modeling expertise. However, Learning from Demonstration (LfD) techniques offer the possibility for the regular Soldier, with no prior coding or ML experience, to customize training by demonstrating the desired behavior. To show the utility of these techniques for autonomous vehicles, a simulated Warthog was trained to perform a military maneuver known as a “berm drill” at Grace’s Quarters Robotics Research Collaboration Campus (R2C2). Data was collected from both a simulated (Unity game engine) environment and at R2C2 during February and March 2025. The final project was presented as part of the successful AIMM-HAT ICE demonstration on 11 September 2025.