Throughput Enhancement Based on Optimized Regularized Particle Filter for H-CRAN

• Published in: IEEE transactions on vehicular technology Vol. 68; no. 6; pp. 5672 - 5683
• Main Authors: Odwa, Lilly-Abau Yosia, Chen, Yue-Yun
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.06.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Chain mobility; Channel estimation; Computational geometry; Computer simulation; Conditional probability; Convexity; Energy sources; H-CRAN; Importance sampling; Kernel; Kernels; Markov chains; Maximum likelihood estimation; Monte Carlo Markov chain; Optimization; Performance evaluation; Radio transmitters; Real time; Real-time systems; regularized particle filter; Resampling; Resource allocation; Resource management; sequential importance sampling with resampling; Throughput; User satisfaction; Wireless communication systems; Wireless communications
• ISSN: 0018-9545; 1939-9359
• DOI: 10.1109/TVT.2019.2908102

The fifth-generation technology is emerging with various real-time applications that demand high data rate. The number of connected devices is expected to increase. However, spectral and energy resources are limited. Therefore, it is challenging to meet users satisfaction. Thus, the necessity for a structure that can sustain real-time performance of wireless communication systems is pressing. This paper presents an efficient radio resource allocation scheme using regularized particle filter with optimized kernel size and kernel bandwidth, and Monte Carlo Markov chain move step based on metropolis algorithm is used to estimate next state of channel. Channel state is calculated from the temporal channel correlation using maximum likelihood estimation. Throughput for the next user state is predicted using posterior probability with normalized weights. The main aim of this paper is to observe the user throughput. The problem is modeled as throughput maximization subject to channel state. The optimization problem is formulated as a non convex optimization and solved using branch and bound method. It has been shown in the simulation results that the average throughput is increased by <inline-formula><tex-math notation="LaTeX">{\text{5.94}\%}{\text{--52.98}\%}</tex-math></inline-formula> compared to sequential importance sampling with resampling (SISR) method. Therefore, the RPF scheme has proved to be efficient by its remarkable increase in user throughput compared to the SISR particle filter algorithm.