Agent-based real-time simulation of autonomous virtual soldiers engaging with moving targets

Virtual combat simulations provide a cost-effective platform for military training; however, their realism is often constrained by deterministic formulations of engagement outcomes against dynamic targets. This study presents a probabilistic attack model for intelligent virtual soldiers that explicitly accounts for target motion, aiming uncertainty, and intrinsic weapon dispersion. Impact locations are modeled as a bivariate Gaussian random variable whose covariance structure reflects both operator-induced aiming error and ballistic variability. To assess operational effectiveness, the drone is decomposed into subsystem-level zones, and hit probabilities are inferred through spatial integration of the impact distribution, enabling the computation of expected damage rather than binary hit outcomes. The proposed model is implemented within a real-time simulation environment and evaluated across varying target velocities. The results demonstrate a nonlinear degradation of hit probability and subsystem-level damage as target speed increases, highlighting the importance of uncertainty-aware modeling for realistic engagement simulation. The proposed model enhances both the fidelity and analytical interpretability of virtual soldier behavior and provides a useful probabilistic basis for defense modeling and simulation of intelligent combat agents.