Data-Driven Optimal Throughput Analysis for Route Selection in Cognitive Vehicular Networks

To meet the dramatically increasing demands for vehicular communications, cognitive vehicular networks have been proposed to broaden the vehicular communication bandwidth by using cognitive radio technology. Meanwhile, the nationwide Super Wi-Fi project that allows the TV white space frequencies to be used for free, makes the concept of cognitive vehicular networks realistic. Recently, lots of technical issues of cognitive vehicular networks have been studied from the network designers' perspective, e.g., vehicular spectrum sensing and access, applications with different vehicular QoS, etc. Different from the existing works, in this paper, we consider from the vehicular users' perspective by optimizing throughput via route selection in cognitive vehicular networks using TV white space. By employing the attainable data rate as route selection metric, we propose two schemes: instantaneous route selection and long-term route selection. To evaluate the expected data rate on the route, we analyze the cognitive vehicular network throughput under two spectrum sharing models: spectrum overlay and spectrum underlay. In the experiments, we use Google spectrum dataset to estimate the intensity of TV base stations in the United States and evaluate the cognitive vehicular network throughput performance, which shows that the spectrum overlay model is more suitable for most of states in current United States, except New Jersey, Delaware and Utah. Moreover, we conduct a case study regarding the route I-88E and I-90E selection between Cortland and Schenectady in New York State. The traffic intensities and traffic intensity transition probabilities of these two routes are estimated using the real-world traffic volume dataset of New York State. Based on the estimated traffic information, we calculate the attainable instantaneous and long-term data rates of each vehicular user, which shows that route I-88E is preferable to route I-90E in most cases.