Research

Short-packet communication (SPC) has been specified as a key technology for the fifth generation (5G) of wireless networks since it meets the stringent requirements on reliability and latency. SPC supports a wide range of ultra-reliable low-latency communication (URLLC) applications such as intelligent transportation systems, high-speed trains, factory automation, and Internet-of-Things (IoT) networks. Deploying cooperative relay for SPCs, along with wireless energy transfer helps to extend the radio coverage and mitigate the pathloss effect. This leads to the reduction of power consumption of IoT devices while meeting the reliability and latency constraints. This work gives fresh and the first endeavor to study an opportunistic relay selection scheme with comprehensive analyses for the BLER, goodput, and EE of wireless-powered cognitive IoT networks with SPCs, followed by deep learning evaluation.

Link: Short-Packet Communications in Wireless-Powered Cognitive IoT Networks: Performance Analysis and Deep Learning Evaluation", IEEE Transactions on Vehicular Technology, vol. 70, no. 3, Mar. 2021.

Motivated by the real-time application requirements and low latency in 5G network, we develop an efficient relay selection scheme based on deep neural network in cognitive IoT systems, which relies on a regression model to provide high accurate predictions. The proposed approach is an efficient deep learning model in reducing the implementation cost by avoiding the derivation of complex closed-form expressions for the system throughput. In this paper, we design a DNN to learn the input-output relation in a general wireless-powered cognitive IoT system to estimate the throughput with high accuracy and low execution time. The resulting trained DNN model can be used as a mapping function to compute real-time performance metrics reflecting the system status. As a consequence, the computational complexity is shifted to the offline training stage, and the complexity during the online prediction stage is significantly reduced.

Link: A Deep Neural Network-based Relay Selection Scheme in Wireless-Powered Cognitive IoT Networks", IEEE Internet of Things Journal, vol. 8, no. 9, pp. 7423-7436, May 2021.

Recently, non-orthogonal multiple access (NOMA) has recognized as the promising multiple access technique for the evolving 5G of cellular networks since it provides high spectral efficiency (SE) and accommodates a large number of users. Simultaneous wireless information and power transfer (SWIPT) has attracted significant attention due to its promising feature in extending the lifetime of wireless networks. Motivated by the 5G network requirements and the benefits of MISO, NOMA, and SWIPT, these concepts can be naturally linked together to realize an efficient type of network model with the SE and energy efficiency (EE) enhancements. In this work, we propose a novel hybrid user pairing beamforming (HUP) scheme for MU-MISO-NOMA downlink systems with SWIPT in order to maximize the achievable spectral and energy efficiencies.

Link: Hybrid User Pairing for Spectral and Energy Efficiencies in Multiuser MISO-NOMA Networks with SWIPT", IEEE Transactions on Communications, vol. 68, no. 8, pp. 4874 - 4890, Aug. 2020.