Abstract:

Select an optimized action policy can make task coalition evolve into the expected operation effect. Considering the influence strength of events is not consistent in different time horizon, timed varied dynamic influence nets are utilized to model optimized action policy selection problem in the process of task coalition evolution. The consistent conditions of probability propagation parameter design are given and the influence constant is computed, both of which are based on causal strength logic. An learnable genetic algorithm (LGA) based on the gene floating theory is designed to solve the optimized action policy selection model. In LGA, in order to enhance the algorithm convergence rate, the whole chromosomes learn the superiority gene bit from the first rank chromosome and combine effective genetic and selecting operators. At last, the simulated results of aerial attack campaign show that optimized action policy selection with various influence constants can improve the capability of cause and effect modeling, and the learnable genetic algorithm good fine convergent and optimizing capability.