A novel advanced 3D‐IPS based on mmWaves and SOM‐MLP neural network

AbstractAbstractThis article proposes a new approach for intelligent indoor 3D positioning that works in the millimeter waves (that constitutes a cornerstone of 5G networks). The solution is based on a combination of a multilayer perceptron network and a self‐organizing map. The solution relies on a precise model of the millimeter frequency channel at 60 GHz and the building of a database of fingerprinting of localization data. The solution approach is based on two parts. In the first part, we specified the target environment for positioning as well as the set of millimeter wave (mmWave) access points (APs) to be deployed in predefined positions. Then, we built a reference database of received signal strengths, azimuths of arrival, and elevations of arrival at each 3D position of the target environment (with a predefined step). The second part of the approach consists of designing a novel neural network architecture that is first trained using the database data and then used to estimate any position in the environment based on the information of the received mmWave signals sent by the different mmWave installed APs. In a normal scenario (without loss of data), the obtained results indicate a millimeter level of the generalization error of the estimated 3D positions. In the case of missing data that may due to potential failures in the system, the obtained accuracy is of the order of 34 cm.This article designs a 3D indoor fingerprinting localization system by using millimeter wave frequency band. With the combined fingerprints of received strength signal, azimuths of arrival, and elevations of arrival at each reference position, an intelligent self‐organizing map neuro‐classifier associated with a set of multilayer perceptrons, achieving a millimeter level positioning accuracy. View Figure