Design and Implementation Artificial Indoor positioning System Based on ANN with PSO Optimization algorithm
Abstract
In this paper, one floor of a specific building was selected to implement Indoor Positioning System (IPS) experiment. To accomplish this work, 4 access points (APs) were installed in the region tested and were located using the "Ekahau Site Survey" to ensure the entire building was covered wirelessly. In addition, a database was created by choosing 58 reference points (RP) in the work region, which included received signal strength (RSSs) collected from all directions for each RP, and these values are recorded using "Net Surveyor 0.2 Package". The Feed Forward Neural Network (FFNN) consists of four layers and uses this network together with the Back Propagation Algorithm (BP) to serve as a localization algorithm for estimating user location. By selecting the optimum number of neurons in each hidden layer and the optimum value of the learning rate using the particle swarming optimization (PSO) algorithm the proposed algorithm is improved. The results showed that the proposed algorithm is characterized by a very short execution time and high accuracy performance