Comparative Study of KPower Means, Ant Colony Optimization, Kernel Power Density-based Estimation, and Gaussian Mixture Model for Wireless Propagation Multipath Clustering

Abstract

In radio communications, channel modeling has a very significant impact. Most of the system's performance relies on the behavior and characteristics of a radio channel. Due to its mobility features, radio channels are time-variant that change over time, making channel characterizations to be dynamic and very challenging. To ensure maximum data rate and reliable communication, accurate channel models are necessary, which requires a correct grouping of wireless multipaths. Accurate clustering of wireless multipath components (MPCs) is essential for cluster-based channel modeling resulting in a more reliable wireless channel system. Currently, determining the best clustering technique is still a challenge as there is no standard way of evaluating and comparing the performance of the various clustering algorithms. This work presents the comparative study on the accuracy performance of the four clustering algorithms namely K-Power Means (KPM), Ant Colony Optimization (ACO), Kernel Power Density-based Estimation (KPD), and Gaussian Mixture Model (GMM) in grouping the wireless MPCs using datasets generated from COST 2100 channel model (C2CM) which represent different Indoor and Semi-urban channel scenarios. Using the Jaccard index, an evaluation metric that compares the calculated data with the reference data, the accuracy of each algorithm is determined as well as their corresponding computational duration. A comparison of their performance is presented, and results show that KPM outperforms other clustering techniques in all channel scenarios of Indoor and Semi-urban environments, making it a right candidate for further improvements to develop a more accurate and computationally efficient clustering technique for wireless propagation multipaths.

Indoor Wireless Multipaths Outlier Detection and Clustering

“…Moreover, KPM keeps its simplicity while reducing the likelihood of being bound by local minima. Using majorization-minimization (MM), power k-means anneals its way to the solution of ordinary k-means, which shares the appealing decent property [10].…”

“…The decreasing amplitude and timing of arrivals of each MPC were integrated into biological ant colony clustering in the Ant Colony Optimization (ACO) algorithm. Each resolvable trajectory component was first projected into a two-dimensional amplitude time plane and then represented as a virtual ant-agent that may move around in this two-dimensional workspace with a preference for high local-environment similarity [10].…”

“…To better identify the local density changes of MPCs, it only used the k-nearest MPCs in the density estimation. To improve the performance of the clustering algorithm, a systematic approach of cluster merging was employed [10].…”

A Novel Approach to Comparative Analysis and Performance Evaluation of Energy Aware Multi-Layered Job Scheduling in Heterogeneous Cloud

“…The position of the nectar food source provides an optimal solution for which the quality of solution is most precise. At the onset; scattered food positions are created where the food source is randomly distributed [16]. The search operations are carried by the scout bees, the employed ones and the onlookers as well; for locating the nectar.…”

“…The ultimate aim of any evaluation system is to keep CPU in busy mode. This parameter is responsible for ascertaining the time used in job execution [16] and also determines the quality feature of the jobs allocated to available resources. Mathematically it is represented by the following expression: Makespan = max (rtj); where the ready time of every scheduling resource is denoted by the term "rtj".…”