An Improved ABC Algorithm Based on Deep Reinforcement Learning for Multi-UAV Target Assignment

The target assignment in multi-unmanned aerial vehicle (multi-UAV) cooperative reconnaissance is a classic problem in weapon-target assignment. Despite the significance of the problem, most of the existing algorithms can't meet the application requirements in solution quality and computational efficiency. Therefore, a self-learning artificial bee colony (ABC) algorithm based on deep reinforcement learning (DRL) is proposed in this study to solve the target assignment problem in multi-UAV cooperative reconnaissance (named DRLABC). In DRLABC, the search equation for the employed bee phase is intelligently adjusted by proximal policy optimization (PPO). Moreover, an improved strategy is adopted for the onlooker bee phase to enhance the comprehensive performance of the algorithm. The learning performance and effectiveness of DRLABC are compared with other rival algorithms using multiple simulation instances with different problem scales. Experimental results show that the proposed algorithm significantly outperforms its competitors in solving target assignment problems.