In the choice process of optimal military commu-nication (MC) alternative, evaluation data mainly come from expert judgments, simulation results and test bed data, and they cannot be directly used in evaluation because of differences in form and attribute; and the MC environment changes rapidly as the operation tempo increasing. It is an important effort to judge the effectiveness robustness of MC alternative, since both the evaluation data and the MC envi-ronment are full of uncertainty. A robustness evaluation method based on multiple data sources and Monte Carlo simluation is proposed with respect to the characteristics of them. Mainly include Belief map as data expression form; Regression relational model built with Support Vector Re-gression (SVR) to acquire simulation data’s confidence with test bed data as training example; Extensive Bayesian Algo-rithm (EBA) to fuse data from multiple sources; Beta distri-bution fitting method for each criterion of each alternative by using the fused results; and calculation of the Probability of Best (PoB) of each alternative through Monte Carlo simu-lation. Take MCE evaluation of a Naval Vessels Fleet as an example, the proposed method is compared with some gen-eral methods. The results indicate that the proposed method helps to obtain relatively conservative alternative and is effective in guaranteeing the robustness.

A military decision maker is typically confronted by the task of determining optimal course of action under some constraints in complex uncertain situation. Thus, a new class of Combinational Constraint Optimization Problem (CCOP) is formalized, that is utilized to solve this complex Operation Optimization Problem. The object function of CCOP is modeled by Influence net, and the constraints of CCOP relate to resource and collaboration. These constraints are expressed by Pseudo-Boolean and Boolean constraints. Thus CCOP holds a complex mathematical configuration, which is expressed as a 0 1 integer optimization problem with compositional constraints and unobvious optimal object function. A novel method of Genetic Algorithm (GA) combination of Boolean Constraint Programming (BCP) is proposed to solve CCOP. The constraints of CCOP can be easily reduced and transformed into Disjunctive Normal Form (DNF) by BCP. The DNF representation then can be used to drive GA so as to solve CCOP. Finally, a numerical experiment is given to demonstrate the effectiveness of above method.