Genetic Fuzzy Tree Based Learning Algorithm Toward the Weapon-Target Assignment Problem

Abstract

Weapon-target assignment (WTA) is essential ability for command and control (C2) systems. The requirement for real-time decision-making, heterogeneous combat platforms are required to make effective weapon-target assignment decisions to achieve interception of fast and multi-batch targets. Since it is difficult to form an accurate modeling of the incoming target ability and obtain a large amount of training data in actual combat exercises, this problem has become a representative problem of real-time decision-making under the constraints of small training samples. Inspired by the use of rules to make coordinated air defense decisions when manned, we propose a practical rule-based machine learning approach to solve this problem in this paper. Firstly, we model heterogeneous combat platforms into multi-agents system and use genetic fuzzy trees (GFT) to make weapon-target assignment decisions. Genetic algorithm (GA) is then employed to learn fuzzy rules and tune membership functions. To evaluate the performance of the proposed algorithm, we build a typical Surface Unmanned System air defense simulation scenario that employs an auto-fire strategy as baseline. The simulation results show that our approach demonstrates a superior performance over the auto-fire strategy and can greatly improve the interception efficiency with a small amount of training data.