Probabilistic/Possibilistic Approach to Modeling C3 Systems. Part 2

摘要

This paper continues the work begun in the last Proceedings (9th MIT/IBR Workshop on Command Control and Communications (C3) system). In that work, C systems are considered as interacting networks of decision-making node complexes characterized by system or process variables. Internodal relations are modeled through nonlinear additive (in the general sense) regression relations; intranodal relations are made to follow a general SGHOR (Sense-Hypothesize-Option-Response) paradigm. In turn, it is shown that a collection of ten types of relatively primitive implication or conditional relations PRIM between C3 variables for enemy and friendly component systems determines all updated marginal node state (distributions. (Distributions can be in the classical probabilistic sense or more generally in a multi-valued logical sense). This leads to a C3 decision game, where the loss function in some picked combination of measures of performance or effectiveness derived from node states and where each decision strategy corresponds to some choice of PRIM for each C3 system. In this work, emphasis is placed upon model refinement. In particular, the intranodal relation representing data fusion is expanded and analyzed. This expansion is characterized by a weighted sum of products for the classical probability case and extended to a more general form for multi-valued logics.