Q-Learning Driven Artificial Bee Colony Algorithm for Solving the Weapon-Target Assignment Problem

Abstract

The Weapon-Target Assignment (WTA) problem is a critical, NP-complete combinatorial optimization problem in military operations research. It aims to allocate defense weapons to incoming targets to minimize the total expected damage to protected assets. This study proposes a Q-learning-driven artificial bee colony (QABC) algorithm that incorporates an adaptive operator selection mechanism based on Q-learning into the standard artificial bee colony (ABC) algorithm. Five permutation-based neighborhood operators are employed to balance exploration and exploitation during the search. We evaluated the proposed algorithm on twelve benchmark WTA instances with problem sizes ranging from five to 200 weapons and targets across 30 independent runs. The experimental results demonstrate that the QABC algorithm consistently outperforms the state-of-the-art method on nine of the twelve instances, achieving substantially lower cost values and standard deviations, especially for large-scale problems.