Spectrum Access Allocation in Vehicular Networks With Intermittently Interrupted Channels

Abstract

In this letter, we study the spectrum access problem in vehicular networks, where a base station (BS) assigns its spectrum to vehicles to fulfill data demands of these vehicles. In our model, connection links between the BS and vehicles are assumed to be intermittently interrupted by local jammers that have attacking strategies following a Markov chain. Furthermore, the channel availability of vehicles is correlated due to the correlated jamming pattern created by different groups of jammers. Consequently, uncertainties in the system dynamics make the spectrum allocation problem in vehicular networks partially observable. Besides, the constraints on data demands, along with the high mobility of the vehicular environment further complicate the design of the access policy. To address the aforementioned issues, the deep Q-learning method is proposed to provide an efficient and structured solution to such spectrum access problem. Besides, the double Q learning method is also integrated into the deep Q-network to improve the training speed of the proposed method. Numerical results are presented to demonstrate the advantages of the proposed policy compared to other benchmark strategies.