Tuning of Empirical Radio Propagation Models Effect of Location Accuracy

The paper summarises tuning of empirical propagation model RX-level predictions to measurements. An algorithm using least-squares fit of predictions to measurements is discussed. The algorithm uses the Condition number of the resulting equations to control the numerical stability of the calculations. The algorithm has been evaluated using measurements obtained in a drive-test measurement campaign in Helsinki. The results indicate that the algorithm performs well by being able to improve the fit of RX-level predictions to the observations. An aim of the CELLO project has been to evaluate the feasibility of using RX-level location data logged by the network to tune the propagation model. If this is feasible, predicted coverage could be continuously re-tuned based on location data acquired during the normal operation of the network. However, because the location methods that will most likely be available in the network are less accurate than GPS-locations, it is necessary to evaluate the effect of less accurate locations on the tuning. The paper thus evaluates the effect of location accuracy on the tuning results. The evaluation is based on measurements where locations were estimated by different methods. A tentative conclusion is made that the location accuracy provided by the DCM (Database Correlation Method) positioning method [H. Jormakka (ed.), S. Horsmanheimo, J. Lähteenmäki, J. Rissanen, Juuso Pajunen, E. Aarnes, S. Holm, B. Forsberg, “Trial Results” CELLO-WP3-VTT-D30, 2003.] could be sufficient for tuning purposes.