Q-learning-driven online target assignment for multi-UAVs considering FAHP situation evaluation

This paper investigates dynamic target assignment (DTA) for multi-UAV cooperative multi-target engagement problem (MMEP) under time-critical and time-varying battlefield situations. A closed-loop framework is developed by integrating a fuzzy analytic hierarchy process (FAHP) based situation evaluation to quantify the UAV–target situation function, and an online Q-learning based assignment module whose reward matrix is updated according to situation variation to enable fast reassignment when the battlefield situation changes. In this case, the R-matrix functions as a battlefield situation monitor to judge the trigger time of target assignment and quantify the number of participants on both sides. Combined with the offline knowledge base obtained by Q-learning, the proposed algorithm can be deployed for online application. Simulations with eight UAVs and four targets demonstrate that the proposed approach achieves reliable optimal assignment within a short time, outperforming a PSO-based baseline and a previous Q-learning scheme.