| We empirically characterize the 5 GHz channel for airport surface (AS) area and vehicle to vehicle (VTV) communication systems. The characterization consists of stochastic models for the channel impulse response, which focus on small-scale, and "medium," or "meso-"scale effects. Motivation is provided by reviewing the growth in civil aviation and VTV communications, and by describing the utility of the 5 GHz band for these new communication systems. Further motivation arose from our literature survey, which revealed a pressing need for wideband stochastic channel models for these new applications in this band. Data measurement campaigns and environment descriptions are provided. For both the AS and VTV settings, classification schemes are developed. These schemes allow grouping AS and VTV environments into classes, and these classes are further divided into propagation regions, for which channel characteristics are statistically similar. A pre-processing framework to extract the most pertinent channel information from the measured data was developed. A propagation path loss model was also developed for the large airport class. Based upon our measured data, we deduced some unique propagation effects: severe, or "worse than Rayleigh," fading, correlated scattering, and statistical non-stationarity (NS). To explain the severe fading phenomenon, we present two physical models that yield results in agreement with our measured data. For each propagation region of the AS and VTV classes, three different small scale fading models (denoted M1, M2, and M3) were developed. These models are applicable to different values of channel bandwidth, allow tradeoffs between the model's implementation complexity and fidelity, and allow for the model user's incorporation of statistical non-stationarity. Channel non-stationarity was modeled using two different random processes: the multipath component persistence process (for AS and VTV) models the finite lifetime associated with a multipath component, whereas the region persistence process (only AS) emulates the transition of the receiver from one propagation region to another. Each of these processes is based on a first-order Markov chain. The channel models were implemented in software using a new correlated multivariate Weibull random process generator. The model outputs were compared with the actual data using both time and frequency domain measures. Our NS model yields best agreement with the data, for all cases. We also present channel models for the cases when the AS transmitter is located at an airport field site. |
