Learn to Coloring: Fast Response to Perturbation in UAV-Assisted Disaster Relief Networks
Bowen Wang, Yanjing Sun, Nan Zhao, Guan Gui
Abstract
In this paper, we address the unmanned aerial vehicle (UAV)-assisted disaster relief networks, where UAVs can disseminate the emergency information to those terrestrial users in a multicast manner. Due to the limitation of spectrum resources, multiple UAVs have to reuse the same channel while the co-channel interference management is needed. The dynamic topology structure induced by mobility can be modeled as a dynamic graph in which the existence of an edge (interference relationship) between two vertices (multicast clusters) is dynamically changing. As such, the channel selection problem can be transformed into a dynamic graph coloring problem in which the graph structure evolves in continuous time slots and the previous coloring strategy becomes valid. In this regard, we propose a stochatic learning based algorithm which can converge rapidly. Simulation results demonstrate that our proposed method is fast in response to the perturbations in dynamic environments.