Particle filter based tracking in a detection sparse discrete event simulation environment

Abstract

One of the key abilities of agents in military simulations is to react to both detections of and counter-detections by other agents in the environment. While methods have been developed to model these detections and counter-detections, the majority of these methods model detection and counter-detection as an all or nothing prospect in which an un-detected entity at some point crosses an arbitrary threshold of observability and becomes fully detected. In actuality, even extremely uncertain or incomplete detections and counter-detections of opposing entities can provide enough data for entities to make reasonably intelligent decisions on the virtual battlefield. Recent developments in commercial gaming artificial intelligence suggest that particle-based tracking techniques can provide accurate and computationally efficient state estimation of opposing agents within virtual environments. In this work several particle-based methods for obtaining and tracking contacts are explored to determine the feasibility of their use as a general purpose tracking technique in military simulations.