Mobility Prediction Optimization of Mobile Hosts in Smart Antennas System Using Adaptive Neuro-Fuzzy Inference System

Abstract

As the demand for wireless communication increases, there is need for better coverage, improved capacity, and higher transmission quality, all of which contribute to better Quality of Service (QoS). One of the promising technologies in achieving excellent QoS is the use of smart antenna systems (SASs) that dynamically radiate power beams to mobile nodes (MNs) in response to received signals to access a wireless link through a process known as beam forming. This has the effect of enhancing the performance characteristics (such as capacity and hand-over) in wireless systems. By using machine learning methods, it is possible to predict the upcoming change in the mobile location at an early stage and then carry out beam forming optimization to alleviate the reduction in network performance. This implies that with a dynamic SAS, a mobile user can be served relatively well while on the move. Efficient prediction of the position of mobile hosts in wireless networks by SASs requires an effective mobility optimization technique. The use static samples of Received Signal Strength (RSS) in locating MNs has also been proposed in many research studies with positive results. This implies that prediction of RSS in wireless networks would form a strong base for mobility prediction and localization. However, these predictions are still challenging issues, which called for this research study. One of the prediction techniques that has been proposed and used is the Grey prediction model (GM) which is associated with benefits of reduced overheads which is a serious issue in wireless cellular networks. This is due to its ability to perform prediction with little data and thus perform with little processing effort. In this research v we used of Adaptive Neuro-Fuzzy Inference System (ANFIS) to achieve better estimations of mobility than the prediction made by conventional models like Log-Normal Shadowing Model (LNSM) and GM. The mobility, in this study, was based on the RSS at the mobile node (MN) as it traverses towards or away from the transmitting antenna. This methodology performs prediction with a mean absolute error (MAE); between 0.0826m and 0.6904m in short distances, and between 0.3220m and 3.8765m in long distances which makes ANFIS one of the excellent methods that have been researched about to solve the mobility prediction issue. The study has also revealed that the average distance at which anomalies in the accuracy of mobility prediction occurs has been identified at 62.33% and 64.82% for short (1m to 100m) and long (100m to 1800m) distance communication environments respectively.