Study On Joint Fire Strike Plan Optimization Problem Based On Stochastic Timed Influence Net

Equipments are the physical foundation of military operation. The physical foundation of Joint Fire Strike (JFS) is the information-based equipments that embed many information technologies especially electronics and computer. The primary characteristic of JFS is that multi-strike forces enable to accomplish"multidimensional integration and compact cooperation"based on C4ISR, which enable JFS to achieve the overall effectiveness. The key process of Joint Fire Strike Planning is the decision of optimal Joint Fire Strike Plan, which includes target selection, fire resource assignment and executing time, and so on.Because of the complexity of situation, Joint Fire Strike Plan Optimization Problem (JFSPOP) contains many dynamic and uncertain complex causal relationships between course of action and effect. Traditional methods such as analytic model and simulation optimization are lack of causal modeling and efficient execution, which are all incapable to solve the JFSPOP. Therefore we propose Possibility Networks (PNs) to solve JFSPOP. However, the traditional PNs cannot fully describe those complex causal relationships, thus we propose Stochastic Timed Influence Nets (STINs) based on PNs combination of stochastic sequence, and then a methodology based on STINs is present to support JFSPOP that enhance the efficiency of planning, and finally guide the warfare. The main work and innovation of this thesis include:(1) The basic mathematic model for JFSPOP and the general solving framework are presented. The concepts relevant to JFSPOP are defined and the mathematical model of JFSPOP is established. JFSPOP is a constraint optimization problem with unobvious optimal object and heterogeneous variable coupling that comprises logistic and numerical constraints. Aiming at the needs of PNs used to model complex and dynamic causal relationship within object function of JFSPOP, a novel JFSPOP solving framework based on the STINs is proposed. This framework should provide the methodology guide for JFSPOP and promote the improvement of efficiency and quality of planning. The methodology is making up of a basic theory named STINs referenced above and tow techniques. The techniques include the object function modeling based on STINs and the optimal solution finding based on Collaboration Evolution Algorithm with Alterable Solution Space (CEA-ASS). Finally, the main research approach of the methodology above is addressed.(2) The effect and delay of operation are usually uncertain for that are subject to status of the nature condition or the aerial defense of enemy etc. Timed Influence Nets (TINs) are a powerful formalism used to model uncertain causal relationship. However, TINs did not meet all requirement of this thesis due to the potential assumption about constant delay. Aiming at the needs of causal modeling, STINs are presented to enhance TINs formalism by relaxing the constant delay assumption, which are used to model dynamic uncertain causal relationship holding stochastic delay. Three key parameters are introduced in STINs that include stochastic delay, delay-varying influences and stochastic belief sequence, and then STINs are formalized mathematically. The basic approach of STINs is addressed based on relational method, which includes parameter learning algorithm, probability propagation algorithm and causal tracing algorithm. Finally, an example is presented to illustrate the feasibility of probability propagation algorithm and causal tracing algorithm.(3) The modeling of object function is the hard core of JFSPOP, the principal work of which is STINs modeling that is enabled to describe the causal relationship among tasks and effects. Firstly, a common process used to model STINs is addressed for supporting JFSPOP, in which the approach capturing parameters of STINs is studied in detail. Then the criterions of alternative assessment are presented based on the character and concept of stochastic belief sequence, which support statistical, confidence and uncertain analysis. Further, the meanings and algorithms of these criterions are explained in full. Finally, the object function of JFSAOP is formalized mathematically based on multi-attribute utility function.(4) Aiming at the needs of solving JFSPOP, a novel CEA-ASS is presented to find the optimal solution of JFSPOP based on the idea of two-level decomposable optimization, in which the constraints of JFSPOP are treated with constraint programming technique. The JFSPOP can be transformed into Static Constraint Optimization Problem (BCOP) and Temporal Constraint Optimization Problem (TCOP) easily. The basic collaboration mechanism of CEA-ASS is solving BCOP and TCOP alternately, the validity of which is proved mathematically. Finally, a novel Genetic Algorithm with binary code based on Boolean constraint programming is addressed for solving BCOP, which translate solution space into Boolean function constraint expressed by Disjunctive Normal Form. A novel Bayesian Optimization Algorithm with integral code based on temporal constraint programming is also addressed for solving TCOP, which translates solution space into full probability function constraint expressed by Bayesian Network.(5) Aiming at the needs of application, a prototype system is achieved based on the methodology mentioned above. The process of the proposed methodology is given and the validity is proven with a case study of aircraft and missile joint strike against aerial dominance.