A Deep Learning-Based Approach to Power Minimization in Multi-Carrier NOMA With SWIPT

Abstract

Simultaneous wireless information and power transfer (SWIPT) and multi-carrier non-orthogonal multiple access (MC-NOMA) are promising technologies for future fifth generation and beyond wireless networks due to their potential capabilities in energy-efficient and spectrum-efficient system designs, respectively. In this paper, the joint downlink resource allocation problem for a SWIPT-enabled MC-NOMA system with time switching-based receivers is investigated, where pattern division multiple access (PDMA) technique is employed. We focus on minimizing the total transmit power of the system while satisfying the quality-of-service requirements of each user in terms of data rate and harvested power. The corresponding optimization problem is a non-convex and a mixed integer programming problem which is difficult to solve. Different from the conventional iterative searching-based algorithms, we propose an efficient deep learning-based approach to determine an approximated optimal solution. Specifically, we employ a typical class of deep learning model, namely, deep belief network (DBN), where the detailed procedure of the developed approach consists of three parts, i.e., data preparation, training, and running. The simulation results demonstrate that the proposed DBN-based approach can achieve similar performance of power consumption to the exhaustive search method. Furthermore, the results also confirm that MC-NOMA with PDMA outperforms MC-NOMA with sparse code multiple access, single-carrier non-orthogonal multiple access, and orthogonal frequency division multiple access in terms of power consumption in SWIPT-enabled systems. INDEX TERMS Non-orthogonal multiple access (NOMA), simultaneous wireless information and power transfer (SWIPT), machine learning.

A New Green Prospective of Non-orthogonal Multiple Access (NOMA) for 5G

“…It is identified as a performance enhancement tool which can achieve intelligence in network operations, optimization, maintenance and management in the emerging mobile networks [159][160][161]. However, there are limited research work [105,162] on improving the energy efficiency of NOMA based networks using branches of AI which leaves space to explore more solutions. In the paper [105], authors have addressed an optimization of energy efficiency and RF energy harvesting using a machine learning based algorithm.…”

“…The combination of BEEM-NOMA and machine learning outperforms the conventional NOMA in terms of energy efficiency. Moreover, in contrast to the conventional methods of optimizing a power allocation algorithm, a deep learning based approach to determine an approximated optimal solution is utilized in [162]. The results of the proposed scheme have shown to achieve a similar performance to the conventional optimization method.…”

Resource Allocation for MC MISO-NOMA SWIPT-Enabled HetNets With Non-Linear Energy Harvesting

“…The authors aimed to optimize the harvested energy and total transmission rate, simultaneously. The authors in [28] investigated the joint resource allocation problem for a TS-SWIPT multi carrier (MC)-NOMA system, where pattern division multiple access (PDMA) technique was employed to minimize the total transmit power of the system. An optimal resource allocation for the MC-NOMA enabled SWIPT technique was proposed in [29] to maximize energy harvesting.…”

Complex-BP-Neural-Network-based Hybrid Precoding for Millimeter Wave Multiuser Massive MIMO Systems

“…Then forward propagation is performed to obtain the corresponding network output and the cost function. Afterwards, the partial derivatives are calculated according to (17), (18), (19) and (20) to adjust the weight. The training process is terminated when the error Algorithm 1 Hybrid precoding algorithm based on BP neural network.…”

“…Calculate the gradient ∇ w (l) nm e 2 according to (16), (17), (18), (19) and (20); 7: Perform the back propagation via SGD and update the weights according to (12), (13) and (14); 8: Calculate the error of the test set. If the error is smaller than the threshold, skip to step 10; 9: end while 10: return Optimized hybrid precoding neural network F. falls to an acceptable range.…”