Measurement-to-track association for nontraditional measurements

Data fusion algorithms must typically address not only kinematic issues - that is, target tracking - but also nonkinematics - for example, target identification, threat estimation, intent assessment, etc. Whereas kinematics involves traditional measurements such as radar detections, nonkinematics typically involves non-traditional measurements such as quantized data, attributes, features, natural-language statements, and inference rules. The kinematic vs. nonkinematic chasm is often bridged by grafting some expert-system approach (fuzzy logic, Dempster-Shafer, rule-based inference) into a single- or multi-hypothesis multitarget tracking algorithm, using ad hoc methods. The purpose of this paper is to show that conventional measurement-to-track association theory can be directly extended to nontraditional measurements in a Bayesian manner. Concepts such as association likelihood, association distance, hypothesis probability, and global nearest-neighbor distance are defined, and explicit formulas are derived for specific kinds of nontraditional evidence.