A Self-Adaptive Artificial Bee Colony Algorithm Based on Generative Adversarial Imitation Learning for Multi-UAV Target Assignment

The target assignment is a crucial task in multiunmanned aerial vehicle (Multi-UAV) cooperative missions. However, most of the existing algorithms can't meet the application requirements in terms of solution quality and computational efficiency. Therefore, a self-adaptive artificial bee colony (ABC) algorithm based on generative adversarial imitation learning (GAIL) is proposed in this study to solve the target assignment problem in multi-UAV cooperative missions (named SAABC). In the proposed algorithm, three heavy-tailed distributions are introduced into the employed bee phase to improve its exploration capability. Then, GAIL is used to dynamically select the random factors for the ABC algorithm. Furthermore, an improved mechanism was introduced in the onlooker bee phase to enhance its exploitation ability. Comparative experiments in different scales are conducted between SAABC and other competing algorithms. The experimental results demonstrate that SAABC can achieve higher-quality assignment schemes and improve allocation efficiency compared to other rival algorithms.