Modeling human lay error on stationary targets with the generalized normal distribution

Abstract

A tank gunner’s ability to lay the reticle on the target is a key contributor to the probability of hitting an intended target. This paper presents a model for predicting human lay error from range with a Generalized Normal distribution (GND). The model outputs a lay error distribution as a function of range, characterized by parameters describing the shape, bias, and standard deviation of the distribution. To compare the GND with the Zero Mean Gaussian distribution (ZMG) currently used in the Tank Accuracy Model, 16,352 trials were collected from 49 subjects on a Unity-based tank simulation. The GND model demonstrated a significantly more accurate characterization of lay error than the ZMG model. The empirical results and modeling approach establish a baseline for human performance when evaluating intelligent targeting systems, and provide a more detailed method for describing lay error distributions.