Universal Signal Propagation Prediction Model Based on PSO-Trained Modified ANFIS for Wireless Communication Networks

Abstract

Currently, it is evident that the use of wireless communication systems is growing at an unprecedented rate. Because of this fact, there is need for accelerated studies on these systems to improve the quality of service (QoS) provided to their users. This can be done through various methods including signal propagation modelling. Correspondingly, the current trend in signal prediction modelling is shifting from empirical and deterministic models to computational intelligence models due to their low computation cost and high accuracy. In this study we have developed a universal theoretical signal propagation model using modified Adaptive Neuro-Fuzzy Inference Systems (ANFIS) trained with Particle Swarm Optimization (PSO) algorithm. The aim is to develop a model that is more suitable for signal propagation prediction. This is an improvement to the original ANFIS structure for wireless communication propagation modelling. In the process of its development the original ANFIS was modified and together with its training algorithm PSO formulated. This was followed by the development of equivalent theoretical ANFIS based models for existing empirical and deterministic models using ANFIS, LOG10D-ANFIS and LOG10D-PSO-ANFIS. The theoretical models were subsequently combined into one universal model. The root mean square error (RMSE), mean error (ME) and standard deviation (SD) of the predicted signal were used in the process of training and testing the model. From the results, the LOG10D-PSO-ANFIS model has very low values in the range of 10-14 to 10-16 for the three-performance metrics compared to 10-7 for LOG10D-ANFIS and 10-1 to 100 for the original ANFIS. Besides the universal model being accurate, it eliminates the need for many input parameters associated with the individual models. This results in just one input, that is, distance being required. It can also be applied in all environments including indoor, outdoor, urban, suburban and rural setups. For the practical modelling of the behavior of the RSSI, a modified ANFIS based practical model (LOG10D-PSO-R-ANFIS) was also developed, and its results compared to those of other models where its performance was found to be superior.