Simulating spatial cross-correlation in vehicular networks

Wireless channels are defined by the presence and motion of objects between and around the communicating stations. As parts of the environment change, so do the channels between stations that are nearby. While the impact of environmental changes on individual channals has been studies extensively, the spatial autocorrelation across multiple channels, which we will call spatial cross-correlation, has received little attention. These effects are important whenever protocols use multiple channels in real time, such as in multi-hop networks. This paper studies the trade-offs between different ways of simulating spatial channel cross-correlation in the context of vehicular networks. We compare independent stochastic, locally cross-correlated stochastic, and explicitly geometric models in terms of both their complexity and the network-level performance they induce. Our results generally favor the geometric approach. Geometric models have higher precision and lower complexity than cross-correlated stochastic models, although collecting the detailed input needed for geometric models can be expensive. As a result, we propose a hybrid approach that combines geometric and stochastic approaches, depending on whether the impact of physical changes has a major or more minor impact on the channels.