1-4244-1513-06/07/$25.00 ©2007 IEEE An Analysis of a Stochastic Urban Propagation Model Using Ray Tracing Generated Results Carmen Cerasoli, James Dimarogonas, Chad Edwards, William Franklin The MITRE Corporation Eatontown, NJ ABSTRACT A stochastic urban electromagnetic propagation model for non-line-of-sight (NLOS) paths was critically examined by comparing model behavior to ray tracing simulations in four city environments. We focus on the quality of model/data fit and the ability to a priori set model parameters based on city geometry and building materials. The stochastic model was found to fit simulated data well in typical cities. However, relating model parameters to city geometry metrics met with limited success. This is most likely due to the difficulty in characterizing city geometry, and the underlying physics of the model. The complex process of electromagnetic propagation is modeled as a simple one-dimensional random work, leading to diffusion-like behavior. The model does possess utility in its ability to provide easily computed estimates of urban propagation path losses and is an improvement over other empirical models. I. INTRODUCTION Providing robust radio communications within an urban environment for highly mobile forces independent of a fixed infrastructure has always been a challenge for the modern Army. Not only do the propagation characteristics limit achievable ranges, they are difficult to accurately predict and possess high variability. The network designer needs to be able to quantify the ranges and associated variability and offer solutions to provide a robust, well-connected network that can support Military Operations in Urban Terrain (MOUT). The complexities of network modeling create the need for an easily implemented, physics-based urban ground-to-ground propagation model for assessing the networking capabilities and performance of proposed DoD RF communications systems. Such a propagation model would provide quick estimates of signal strengths in generic cities without using computationally intensive techniques such as ray tracing where detailed building structure data (non-generic) are required. This model would not be used to plan a specific deployment with highly accurate predictions on a node-by-node basis, but rather provide a generic assessment of the connectivity expected for the network as a whole. This would provide estimates for the network behavior of a large number of nodes and parameter variations without being so computationally intensive as to create unreasonably long run times. A number of highly empirical models exist for elevated base stations to ground mobile users such as the Okamura-Hata or Walfisch-Ikegami models [1]. Models exist for NLOS, ground-to-ground, paths [2, 3] but are designed for relatively orderly urban building arrangements. Franceschetti and others [4, 5] have recently proposed stochastic models for electromagnetic propagation in urban environments. These models are based on a well defined physical picture where RF propagation is modeled as a one-dimensional random walk. They predict the functional form for received signal strength versus transmitter distance and require at most three parameters: an effective power, a measure of inter- building spacing and a measure of the building reflection characteristics. This paper critically examined the capabilities of stochastic propagation model and attempted to relate city characteristics to model parameters. The analysis relied on ray tracing results generated for four cities with very different geometries: Rosslyn, Ottawa, Berne and Helsinki. Three frequencies were chosen, 400, 900 and 2100 MHz, and building material characteristics were varied. We investigated the ability to set model parameters using city geometry information. Such a capability would allow model prediction by simply computing the appropriate city geometry metric(s) and relating them to model parameters. Section II provides a brief discussion of the ray tracing approximation and defines the simulation strategy. The stochastic urban model is described in Section III, and ray tracing simulation results are presented in Section IV. Model and simulation results are compared in Section V where we attempt to relate city geometry metrics to model parameters. Section VI discusses the models capabilities and short comings and presents two approaches for model