Research On CGF Multi-objective Optimization Decision-making Behavior Modeling Based On Diversity Voting

Multi-objective Optimization is one of important research topics in Decision Making Behavior Modeling of Computer Generated Forces.The computing efficiency of multi-objective optimization algorithms is of vital importance to the believable representation of agent’s decision-making behaviors.Currently however,most MOO techniques used in Game AI or CGF decision-making behavior modeling cannot satisfy the great demand in future simulation system in aspect of solution effectiveness as well as time efficiency.The main reason behind this is that there exists no algorithm that could search and evaluate all possible states,while at the same time the complex multi-objective optimization problem in CGF behavior modeling usually has the feature of large state space,high computing complexity,high frequency of problem solving and strict real-time constraint.According to the above problem,taking the Tactical Position Selection problem in Combat Simulation System as research background,and improvement of the time efficiency of multi-objective optimization problem solving among Combat Simulation as objective,the paper proposes a novel diversity voting model based on Group Diversity and Social Choice theory to solve multi-objective optimization problems.The paper first illustrates the research background and its significance.Further,disadvantages are analyzed and the paper suggests main research contents and generalizes its innovation points.The main problems researched by the paper are the team formation and opinion aggregation methods in the diversity voting model.The existing diverse problem-solving team models lack the quantitative description and effects analysis of diversity itself.Accordingly,the paper proposes detailed mathematical description of team diversity and further gives its effects assumptions,theories and experimental analysis.Besides,existing models usually believe that solution utility is known to each problem solver,thus neglect the fact that problem solvers mostly know nothing about solution utility in real surroundings.Therefore,the paper establishes the voting model of diverse team of problem solvers to aggregate team opinions.In the aspect of solving multi-objective optimization problem using diversity voting model,the paper changes the general model according to the reflecting relationships between MOO problem and diversity voting model.The paper fully considers the feature of multiple Pareto results in MOO problem solving and establishes the diversity voting model based on Approval voting.This improves the flexibility as well as information using efficiency of the model.Considering the time efficiency improvement of diversity model in solving multi-objective optimization problem,the paper generalizes the feature of little interactions among problem solvers in diversity voting model and takes advantage of the parallelization of diversity model.Further,the paper tests the problem solving ability of diversity model under parallelized conditions.The paper takes tactical position selection problem in CGF combat simulation system as research background.It first researches on the problem formalization under multi-objective optimization settings.Further,the paper carries out TPS problem solving based on single MOO algorithm and on Diversity Voting model respectively.It makes comparisons of two approaches according to time and solution efficiency which further testify related assumptions of team diversity itself.At last,the paper concludes its research work and casts prospects into future research as well as focused problems.