A Statistical Approach Toward Channel Modeling With Application to Large-Scale Censored Data

Abstract

Connected vehicle safety and efficiency applications rely on vehicular wireless networks for information sharing. To study the performance of these networks, and numerous protocols designed for them, it is essential to be able to model the channel behavior. However, due to the complexity of the vehicular environment, such modeling has to rely on data collected from large scale field tests. A majority of data collection campaigns with a large number of vehicles utilize vehicular on-board units (transceivers) to collect data in the form of received signal strength (RSS) values. Such datasets are limited in information since only the RSS values of correctly received packets are recorded, leading to datasets that can be considered as censored; especially where packet losses are high. Therefore, deriving channel models that accurately represent the channel behavior at distances with considerable loss of packets is a challenge. In this paper, we propose Bernoulli Mapping Estimation, a novel approach for the estimation of distributions from censored samples. Additionally, we provide a general method and formulation for modeling the communication channel in vehicular environments, considering receiver characteristics, imperfections, and data sparsity. The proposed methods were applied to datasets collected from highway and intersection environments; and the derived models are shown to be accurate representations of the datasets. It is also shown that even with only a small fraction of the RSS values, the framework is able to produce fading models that are similar in characteristics to field data.