Optimization Of Firepower Allocationbased On Wargame

Most of the traditional fire allocation methods focus on the comparison of the forces between enemy and us in the initial situation of the battlefield,and then carry out shell-eye planning.Although these methods get the optimal solution under the constraints of the relevant fire allocation model,it is often difficult to verify these fire allocation schemes.In addition,there are certain limitations in planning the firepower distribution scheme only considering the enemy and our forces in the initial situation of the battlefield.With the advance of the battle situation,ammunition consumption and target damage may no longer be constrained by the corresponding firepower distribution model.Therefore,it is necessary to study the optimization of firepower distribution scheme.Combining with the war game exercise system in solving the static power allocation scheme based on initial battlefield situation,on the basis of using reinforcement learning algorithm in a war game system training agent,to be able to dynamically according to the battlefield new firepower allocation plan for the new situation,thus to optimize implement the original plan,the effect of Finally,an example is taken for evaluation and verification.Therefore,the main research work of this study is as follows:(1)The limitations of existing research methods of firepower allocation are sorted out,the research methods and specific algorithms are determined,and the overall technical framework of firepower allocation optimization and scheme evaluation based on military game deduction is constructed.(2)A fire distribution model with multiple ammunition types and multiple platforms was constructed,and a rule-based pruning algorithm was proposed to solve the problem of large space for this model.In order to solve the problem of large solution space,an improved genetic algorithm is proposed.The main work lies in the adaptive improvement of crossover operator and mutation operator in the algorithm.(3)In the process of solving the problem that genetic algorithm can not dynamically plan and adjust according to battlefield changes,a solving model based on reinforcement learning algorithm is proposed.The reinforcement learning algorithm needs to combine with the environment to train the agent.In this paper,q-learning algorithm and DQN algorithm are used to train the agent,and the corresponding reward value design and dynamic decision method are proposed.Aiming at the deficiency of DQN algorithm,the sample extraction method is extended,and the DQN method based on the best experience playback is introduced.(4)Aiming at simulating the fire attack process with the military chess deduction system,the paper puts forward the development process and key technologies of the military chess system,and introduces the main functions and realization methods of the map module,operator module,rule module,decision module and system interface module.(5)Taking a certain firepower combat as the background,the military game deductive analysis and evaluation verification are carried out with the above algorithms.After pruning the bullet-mesh combination by rule-based pruning algorithm,the static fire distribution scheme is solved by improved genetic algorithm.Q-learning algorithm,DQN algorithm and DQN algorithm based on best experience replay are used to realize dynamic fire allocation planning in military game deduction environment.The case analysis shows that the optimization method of firepower distribution scheme based on military game deduction is reasonable and has good reference significance.