GAN-assisted NSGA-II for UAV scheduling in post-disaster relief scenarios
Guanhao Zhou, Xuzhao Chai, Lijie Ren, Jiajun Ma · 2025
Unmanned Aerial Vehicles (UAVs) are widely used in post-disaster situations because they are flexible, fast, and can reach hard-to-access areas. However, when there is high uncertainty in the disaster environment, it becomes difficult to make good decisions about how to schedule UAVs and distribute resources. To solve this problem, this paper proposes a GAN-assisted NSGA-II optimization method for UAV scheduling under uncertain demand. A robust optimization model is built based on a budgeted uncertainty approach, which considers differences in rescue priority and changes in regional needs. A Generative Adversarial Network (GAN) is added to help the algorithm learn the distribution of good solutions during the search process. This GAN creates new high-quality solutions that help increase variety in the population and avoid getting stuck in poor results. Two simple rules are also designed to sort the UAV paths: one is based on distance, and the other on task priority. Test results show that the proposed method improves the performance of the algorithm in terms of solution diversity, stability, and speed of convergence, even when the uncertainty level changes, identifying optimal perturbation settings for the considered disaster environment.