Skip to Main Content
The database correlation method (DCM) is a network based positioning technology which has shown superior in terms of accuracy. DCM is based on a pre-measured database of a location dependent variable such as received signal strength (RSS). Even though the technique has good potential, the practical difficulty in forming the database (fingerprints) using field measurements has become the major challenge in implementing this in large, dynamic networks. A remedy for this is to make use of propagation model predictions instead of field measurements to create the fingerprints. However, due to the considerable deviation between the predictions and the actual measurements, the positioning accuracy diminishes with this approach. In order to overcome this issue, tuning of the predictions using a small number of field measurements can be applied. The work presented in this paper proposes a technique for the correction of such deviations which would improve the performance of DCM. The proposed tuning process, cell-wise calibration, is based on artificial neural networks (ANN). Two different training algorithms, particle swarm optimization algorithm (PSO) and BFGS algorithm are applied for ANN training. The results of the trials carried out in urban, suburban and rural environments are presented. With the PSO algorithm, the level of accuracy is comparable to that obtained with a measured fingerprint database in urban and suburban environments, and is better in rural environment.