AG Kommunikationstheorie


A Toolbox for Urban Multipath Radio Wave Propagation Prediction


This thesis contributes a concept and its implementation for multipath radio wave propagation (RWP) pre- diction for urban environments. RWP predictions are essential for applications like the Covering-Location- Problem in network planning. Measurement campaigns are too complex and resource intensive and pure stochastic propagation models deliver a very good average performance, however do not cover particular scenarios. Motivated by this, a propagation prediction tool reflecting the particular environment geometry is of high interest. A further imperative is practical applicability, with the intention to be fast and use easily acquirable input data. The environment data and the path loss model play a crucial role. We use so called 2.5-d data, where buildings are not only defined by their outline but also by the ground level and height. Available sources are land-registries or OpenStreetMap, for instance. But before the environment data is used, some pre- processing is mandatory. This yields a set of prerequisite information the simulation relies on blindly for better performance. Since it is infeasible to measure the error imposed by the pre-processing, in terms of the propagation prediction results, we developed a method to quantize the amount of deviation. Hence, we can optimize the processing steps in order to achieve the required result and minimize the error. Propagation paths from the transmitter are modeled by a graph, which forms the basis for the new algorithm in this thesis. Instead of conventional 3-d ray tracing, in two successive 2-d steps first the view from above is processed and thereafter height information is handled. This eliminates redundancy and improves computational speed significantly. After the graph model is completed, concrete multipath information, sequences of coordinates, is in- stantiated. Each coordinate bears additional information about the type of deflection and the surface. The algorithm scales well ranging from calculating paths for a couple of receivers up to millions of receivers. Thus, it may be used to produce a solution for a simple route of a car or pedestrian, or to cover the whole scenario in a grid of receivers. A properly calibrated path loss model mitigates the effects of the incomplete environment data. The suggested calibration process searches a parameter set, minimizing the RMSE. To achieve this, we follow several steps. Initially we estimate the strongest path per receiver. With these paths and the reference data we create an equation system and solve it. This first parameter set is iteratively refined considering new path combinations. So, our tool generates multipath information in minutes for complex scenarios and in sub-second runtime for small scenarios. The output is a set of transmitter receiver links taking deflection effects in a simplified environment model into account. Finally, we can efficiently evaluate that output in the post-processing to various ends. One application could be creating a path loss map as input for network optimization tools. Since the structure allows adding an antenna field pattern in the post-processing, a multitude of configurations can be tested efficiently. For a whole new field of applications, we extend the multipath information and embed it into a MIMO channel model. In a novel semi-stochastic channel model we combine the deterministic output with the geometry-based stochastic WINNER II channel model.

zurück zur Terminübersicht