Research on Intelligent Missile Target Assignment Method Based on Target System Value and Deep Reinforcement Learning

Missile target assignment (MTA) is a difficult problem in the field of military operations research and command and control, and the current mainstream MTA methods still lack efficiency and accuracy in target value assessment and solving large-scale combat scenarios. To this end, we propose an intelligent missile target assignment method (SV-MTA) based on target system value. To solve the deficiencies of system coupling and emergence difficult to quantify in traditional target value research, we design a Material and Informational Flow Cycle Algorithm (MIFCA), which quantifies the importance of targets in the system by calculating the material information flow efficiency of targets in complex networks and provides objective To compensate for the shortcomings of the traditional MTA research in which the confrontation conditions are too simple, the missile attack and defense is regarded as a Markovian decision process of dynamic confrontation, and the concept of target "health" is introduced, and the decay of "health" is used to define The concept of target "health" is introduced, and the decay of "health" is used to define the damage requirement, combining the missile range, surprise defense capability and hit probability to enrich the dynamic scenario of the confrontation between the enemy and us. To avoid the problem of large state space and "dimensional explosion" in traditional reinforcement learning, an improved deep Q-network is constructed to interact with the simulator to solve the confrontation model, which can automatically solve the optimized attack strategy without relying on a priori information. The simulation results show that SV-MTA has the advantage of fast and accurate solutions to large-scale MTA problems compared with traditional methods, and has better robustness to changes in MTA scenario elements, and the optimized strategy can achieve high-quality missile resource allocation and demand prediction.