Inverse Modeling of Combat Behavior with Virtual-Constructive Simulation Training

Abstract

Modeling combat behavior is an important, yet complicated task because the combat behavior emerges from the rationality as well as the irrationality. For instance, when a soldier confronts a dilemma on accomplishing his mission and saving his life, it is difficult to model his ongoing thoughts with a simple model. This paper presents (1) how to reconstruct a realistic combat environment with a virtual-constructive simulation, and (2) how to model such combat behavior with the inverse reinforcement learning. The virtual-constructive simulation is a well-known simulation application for soldier training. Previous works on this virtual-constructive simulation focuses on a small number of entities and mission phases, so it was difficult to observe the frequent behavior dilemma in the field. This work presents a large scale and a complete brigade-level operation to provide such synthetic environment to human player. Then, our second work is observing the com-bat behavior through the virtual-constructive simulations, and modeling the behavior with the inverse reinforcement learning. Surely, we can observe the descriptive statistics of the observed behavior, but the inverse reinforcement learning provides calibrated weights on the valuation on hypothetical rewards from conflicting goals. Our study is the first attempt on merging the large-scale virtual constructive simulation and the inverse reinforcement learning on such massive scale.