Measuring Information Gain in Tactical Operations
暂无分享,去创建一个
Abstract : A measure is proposed for analytically determining the amount of information gained by a tactical battle commander as a result of intelligence, scouting or reconnaissance reports. The measure is based on concepts from information theory, and involves modeling a commander's uncertainty in terms of probability distributions over sets of possible states his adversary may occupy. As the commander gets information, these distributions are updated, by various means, to represent his current state of uncertainty. The information gain is defined in terms of the "distance" between the initial and updated states of uncertainty. Based on a set of plausible assumptions about how an information gain measure should behave, it is shown the measure must be of a certain form involving the decrease in entropy (as defined by Shannon) from the prior to updated distributions. Implications of this characterization are presented and illustrated with examples. Applications to experiments performed at the U.S. Military Academy are described, including: (1) a Janus combat simulation study of the relative reconnaissance performances of the Comanche helicopter and an unmanned aerial reconnaissance system; (2) a Janus-based experiment designed to establish links between the level of information possessed by a combat commander and the degree of success he enjoys against his adversary; (3) a simulation-based design study of intelligent minefields; and (4) development of an information gain MOE for Janus analyses. We believe these examples demonstrate the potential utility of the information gain measure for a wide variety of a applications.
[1] Donald Barr,et al. An indicator of combat success , 1993 .