Human operators in adversary tracking encounters

In the past, human operator models have been developed for tracking tasks by considering that the tracking input signal is passive. A passive signal is an externally applied signal which is generated independently of the tracking loop composed of the operator and his machine. If the signal is an intelligent adversary who reacts to the operator´s actions, then the tracking loop is complicated by the inclusion of the adversary system. This is called the active tracking problem. The objective here is to determine through linear-quadratic-Gaussian differential game theory with perfect state information whether conclusions based upon passive signal analysis is applicable to the complex adversary encounter. We assume that the passive signal is a stored measurement of a previously performed encounter. Although the optimal controls obtained for passive and active tracking are functionally quite different, the controller for the active engagement can be constructed from that of the passive tracking case. It is shown that conclusions based upon the passive encounter do not, generally, predict the active encounter. The tracking error variance and control effort calculated for the passive case are both smaller than that predicted by the active encounter. In practice this is due to the assumed lack of correlation of the input signal with the noise signatures in the passive tracking loop. If the input signal is properly correlated with the tracking loop, the measures of performance in the active and passive tracking loops are coincident.