Performance Analysis of Cooperative Non-Orthogonal Multiple

Access System Using Deep Learning Technique

M. Ramadevi, Motla Kundhan Reddy, Chidurala Saiteja and Tummala Deepak Raja

Department of ECE, VNR VJIET, Bachupally Hyderabad, 500090, India

Keywords: CNOMA, NOMA, Convolutional Neural Network, BER, SNR.

Abstract: Non-Orthogonal Multiple Access (NOMA) has emerged as a promising technique to enhance spectral

efficiency in wireless communication systems. It allows users with varying channel conditions to share the

same frequency band, enabling simultaneous transmission. Cooperative NOMA, an extension of NOMA,

further elevates system performance by leveraging cooperative communication and exploiting spatial

diversity. This study proposes a novel approach for evaluating the performance of cooperative NOMA systems

utilizing deep learning techniques, specifically Convolutional Neural Networks (CNNs). Unlike conventional

analytical methods, deep learning models excel in capturing intricate patterns and correlations in large-scale

wireless communication systems, rendering them well-suited for performance analysis tasks. The suggested

CNN-based framework is trained using simulated data derived from a cooperative NOMA system model.

Inputs to the CNN encompass channel state information, power allocation parameters, and other pertinent

system parameters, while the output entails achievable Outage Probability or Bit Error Rate predictions. By

discerning the relationship between system parameters and performance metrics, the CNN adeptly forecasts

the cooperative NOMA system's performance across diverse scenarios. The effectiveness of the proposed

approach is assessed through extensive simulations, wherein the performance of the CNN-based model is

juxtaposed against traditional analytical methods. Results indicate the CNN's superiority in terms of accuracy

and computational efficiency, particularly in scenarios characterized by complex channel conditions and

dynamic network environments. In summary, this study underscores the potential of deep learning techniques,

particularly CNNs, in scrutinizing and optimizing cooperative NOMA systems. This paves the way for the

streamlined design and deployment of next- generation wireless communication networks, promising

enhanced efficiency and performance.

1 INTRODUCTION

Analyzing the performance of Cooperative Non-

Orthogonal Multiple Access (CNOMA) systems,

with a focus on outage probability and bit error rate

(BER), through the novel utilization of deep learning

techniques such as Convolutional Neural Networks

(CNNs), represents a significant leap in wireless

communication research (Dong, Chen, et al. , 2017).

CNOMA, blending cooperative communication and

Non- Orthogonal Multiple Access (NOMA)

principles, holds immense potential for enhancing

spectral efficiency and system reliability in wireless

networks. In this study, our objective is to leverage

CNNs to conduct an extensive exploration of

CNOMA system performance (Sari, Gui, et al. ,

2018).

Diverging from traditional analytical methods,

which may fall short in capturing the intricate

dynamics of large-scale wireless communication

systems, deep learning models offer an appealing

alternative by discerning complex patterns and

correlations directly from data (Sari, Adachi, et al. ,

2020). Our aim is to train CNNs using simulated

datasets derived from CNOMA system models,

aiming to develop robust predictive models adept at

accurately estimating outage probability and BER

across diverse operational scenarios. To achieve this

goal, we follow a comprehensive research

methodology. We meticulously construct a CNOMA

system model, delineating parameters like user

distributions, channel characteristics, power

allocation strategies, and cooperative relay schemes

(Gui, Song, et al. , 2018). Using simulation tools like

908

Ramadevi, M., Reddy, M. K., Saiteja, C. and Raja, T. D.

Performance Analysis of Cooperative Non-Orthogonal Multiple Access System Using Deep Learning Technique.

DOI: 10.5220/0013734800004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 3, pages 908-913

ISBN: 978-989-758-763-4

MATLAB, we generate synthetic datasets spanning a

wide spectrum of CNOMA system scenarios,

encompassing variations in channel conditions, user

densities, and system configurations (Huang, Guo, et

al. , 2019), (Huang, Guo, et al. , 2019).

These synthetic datasets serve as the training data

for our CNN-based predictive models. Input features

fed into the CNNs include channel state information

(CSI), power allocation parameters, cooperative relay

details, and other relevant system attributes, while the

desired performance metrics— outage probability and

BER—serve as the output (Lu, Cheng, et al. , 2021).

During the training phase, standard deep learning

techniques such as backpropagation and stochastic

gradient descent are employed to fine-tune the CNN

parameters, minimizing the gap between predicted

and actual performance metrics on the training dataset.

Techniques like cross-validation and regularization

are utilized to ensure the generalization and

robustness of the trained models. Subsequently, the

performance of the CNN-based predictive models

undergoes rigorous evaluation using separate

validation datasets, assessing their accuracy and

reliability in predicting outage probability and BER

across a spectrum of CNOMA system scenarios.

Comparative analyses with conventional analytical

methods are conducted to validate the effectiveness

and computational efficiency of the CNN-based

approach. Through this comprehensive analysis, our

research aims to provide profound insights into the

behavior and optimization of CNOMA systems,

particularly concerning outage probability and BER.

The resultant CNN-based predictive models hold the

potential to revolutionize the design and optimization

of CNOMA-based wireless communication

networks, ushering in an era of heightened spectral

efficiency, reliability, and performance in future

wireless systems.

2 OBJECTIVE

2.1 Comparison of Outage probability

of NOMA and CNOMA system

with and without deep learning.

When evaluating wireless communication reliability,

it is crucial to include the outage probability, which

represents the possibility of not meeting quality of

service requirements. Systems like NOMA and

CNOMA provide improved spectrum efficiency and

fairness. Without deep learning, conventional

algorithms have a major role in performance, which

could result in less than ideal results. On the other

hand, incorporating deep learning enables adaptable

parameter changes depending on dynamic

circumstances, enhancing resource distribution and

interference control. Deep learning can, in general,

dramatically lower the likelihood of an outage in both

NOMA and CNOMA systems.

2.2 Comparison of Bit error rate of

NOMA and CNOMA system with

and without deep learning.

An important metric for assessing communication

quality is the bit error rate, or BER. The goal of

NOMA and CNOMA systems is to maximise

dependability and spectrum efficiency. Performance

is dependent on conventional methods in the absence

of deep learning, which might not fully utilise system

capabilities. Adaptive modulation, coding, and

decoding are made possible by integrating deep

learning, which enhances error detection and repair.

In general, BER in both NOMA and CNOMA

systems can be greatly reduced by utilising deep

learning, improving system reliability overall.

3 BLOCK DIAGRAM

Figure 1: CNOMA System

The setup involves a single antenna utilized by

both the source and users due to their significant

distance from each other. In this configuration,

besides serving as a direct transmission link between

the source and both users, In addition, the Near user

acts as a conduit for the Far user. Additionally, the

Near user can harvest energy from the received signal

by using the power splitting (PS) protocol

compensating for energy loss during the relaying

phase. The transmission process can be divided into

cooperative and direct transmission phases. Signal

Performance Analysis of Cooperative Non-Orthogonal Multiple Access System Using Deep Learning Technique

909

detection for the Far User is achieved using a

Successive Interference Cancellation (SIC) method

based on Deep Learning (DL).

Figure 2: CNN Layers

A CNN model learns spatial feature hierarchies

(e.g., horizontal, vertical lines) using the back

propagation algorithm, leveraging building blocks

such as fully linked layers, pooling layers, and

convolutional layers. Typically, the CNN consists of

the following layers: Convolutional Layer: The initial

layer in the CNN, it consists of convolutional kernels

(filters) where each neuron acts as a kernel.

Convolutional pooling is utilized to process multiple

inputs concurrently. Pooling Layer: Following the

convolutional layer, this layer reduces the

dimensionality of feature maps, extracting essential

features while discarding irrelevant information.

Fully Connected Layer (FC layer): Located at the end

of the CNN, These layers serve as a link between each

neuron in one layer and every other layer's neurons,

facilitating complex pattern learning and accurate

predictions

The CNN architecture generally features

alternating convolution and pooling layers, often

concluding with one or more FC layers.

Convolutional layers play a key role in feature

extraction, computing parameters like transmitted bits

Figure 3: Graphical Analysis of BER for NOMA vs

CNOMA:

and channel information. Ultimately, the

classification phase enables precise recognition of

transmitted data.

From Graphical analysis of horizontal green curve

which indicates NOMA BER curve is below the Black

curve which indicates CNOMA BER curve. From the

graph different pairs of BER vs SNR values are

tabulated and compared.

Table 1: Tabular Analysis of BER for NOMA vs CNOMA

The above table determines the comparison of

BER of NOMA and CNOMA systems for a given

SNR values without using Deep Learning. From the

table we can conclude that BER values of NOMA is

higher than that of CNOMA . When the transmitted

power is increasing down the column the BER values

of both NOMA and CNOMA are decreasing.

Figure 4: Graphical Analysis of BER for NOMA vs

CNOMA with DL:

From Graphical analysis of horizontal green curve

which indicates NOMA Ber curve is below the Black

curve which indicates CNOMA Ber curve . From the

graph different pairs of Ber vs SNR values are

tabulated and compared.

INCOFT 2025 - International Conference on Futuristic Technology

910

Table 2: Tabular Analysis of BER for NOMA vs CNOMA

Using DL:

The above table determines the comparison of

BER of NOMA and CNOMA systems for a given

SNR values using Deep Learning technique that is

Convolutional Neural Network. From the table we can

conclude that BER values of NOMA is higher than that

of CNOMA. When the transmitted power is increasing

down the column the BER values of both NOMA and

CNOMA are decreasing.

Figure 5: Outage Probability for NOMA vs CNOMA:

From Graphical analysis we have plotted the

graphs of analytical and simulated outage probability

values of both NOMA and CNOMA systems . Theta

determines the available spectrum and beta

determines required different frequency allocations.

For different beta and theta values outage probability

of Near and Far users of both the systems plotted and

tabulated.

Table 3: Tabular Analysis of Outage Probability for

NOMA vs CNOMA

SNR(DB)

Near

User

(U1)

Far

User(U2)

Near

User(U1)

Far Use

(U2)

0.57

0.61

0.51

0.57

0.31

0.36

0.26

0.29

0.21

0.26

0.12

0.19

0.16

0.17

0.04

0.119

0.13

0.15

0.02

0.114

The above table indicates the comparison of

outage probability of NOMA and CNOMA systems

for a given SNR values without using Deep Learning

technique. When we compare far and near users of

both systems near user outage probability is lesser

thanthe far user, but when we compare the near user of

both the systems and far user of both the systems

CNOMA’s outage probability is lesser than NOMA

due to the realy connection between the users.

Figure 6: Outage Probability for NOMA vs CNOMA using

Deep Learning:

From Graphical analysis we have plotted the

graphs of analytical and simulated outage probability

values of both NOMA and CNOMA systems . Theta

determines the available spectrum and beta

determines required different frequency allocations.

For different beta and theta values outage probability

of Near and Far users of both the systems plotted and

tabulated.

Table 4: Tabular Analysis of Outage Probability for

NOMA vs CNOMA using Deep Learning:

SNR(DB) Near User

(U1)

Far

User(U2)

Near

User

(U2)

Far

User(U1

)

5 0.49 0.58 0.48 0.52

10 0.29 0.32 0.28 0.27

15 0.14 0.23 0.18 0.14

20 0.13 0.19 0.06 0.08

25 0.09 0.18 0.03 0.06

The above table indicates the comparison of

outage probability of NOMA and CNOMA systems

for a given SNR values using Deep Learning

technique i.e Convolutiona Neural Network. When we

compare far and near users of both systems near user

outage probability is lesser than the far user, but when

we compare the near user of both the systems and far

Performance Analysis of Cooperative Non-Orthogonal Multiple Access System Using Deep Learning Technique

911

user of both the systems CNOMA’s outage

probability is lesser than NOMA due to the relay

connection between the two users in CNOMA

system.

Therefore when we compare the values of outage

probability of NOMA and CNOMA systems with and

without Deep learning from both the tables we can

conclude that the outage probabilities of both the

systems is reduced from normal conventional SIC

(Successive interference cancellation) to Deep

Learning based SIC.

4 CONCLUSIONS

We have shown how cooperative NOMA systems

execute performance analysis, using deep learning

techniques to assess Bit Error Rate (BER) and Outage

Probability. Our discoveries have provided important

new information on the potential and capabilities of

these kinds of systems. First off, our findings show

that cooperative NOMA has significant performance

advantages over conventional orthogonal multiple

access systems in terms of BER and outage

probability.

NOMA improves spectrum efficiency and

reliability through effective power allocation and user

grouping, especially in situations with a high user

density and a variety of channel circumstances.

Second, there are encouraging outcomes when

deep learning methods are combined with outage

probability estimation and BER. When compared to

traditional analytical techniques, convolutional neural

networks (CNNs), a deep learning model, provide

higher accuracy and robustness in capturing

complicated channel behaviours and predicting

performance indicators. These models provide more

precise and effective predictions by learning from

large-scale datasets, which helps with resource

allocation and system optimisation.

Furthermore, in order to maximise the

performance advantages of deep learning- based

approaches, our study emphasises the significance of

appropriate model architecture design, training data

selection, and optimisation methodologies. The

models accuracy and generalisation abilities can be

further improved by adjusting network settings,

utilising transfer learning, and investigating

innovative architectures customised to particular

NOMA system features, particularly in dynamic and

heterogeneous wireless environments.

5 FUTURE SCOPE

It is anticipated that the combination of deep learning

techniques and cooperative NOMA systems will

greatly advance the field of wireless communication

research. Investigating cross-layer methods, multi-

objective optimisation, and dynamic resource

allocation has the potential to completely transform

the fairness and efficiency of systems. Furthermore,

extending deep learning frameworks to massive

MIMO environments and heterogeneous networks

presents opportunities to improve system capacity

and spectral efficiency. In order to refine these

systems for practical implementation and sustainable

operation, real-world deployment efforts and an

emphasis on robustness, security, and energy

efficiency will be crucial. When combined,

cooperative NOMA systems and deep learning signal

a new direction in wireless communication research

that should help address the demands of future

connection.

ACKNOWLEDGEMENTS

The author extends sincere thanks to Kundhan, Sai

Teja, Deepak for their valuable contributions in

discussing the results and providing feedback.

REFERENCES

Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou

Tang. “Image super- resolution using deep

convolutional networks”, IEEE Transactions on Pattern

Analysis and Machine Intelligence, arXiv:1501.00092,

38(2):295–307, 31st July2016.

Hikmet Sari, Guan Gui, Hongji Huang, and Yiwei Song,

“Deep learning for a nonorthogonal multiple access

strategy that works well”, in IEEE Transactions on

Vehicular Technology, 67(9):8440–8450, July 2018.

Hikmet Sari, Fumiyuki Adachi, Zhen Yang, Jianhua Zhang,

Song Guo, Guan Gui, and Hongji Huang

“Opportunities, difficulties, and solutions related to

deep learning for physical-layer 5G wireless

technology” , IEEE Wireless

Communications ,Volume: 27, Issue: 1, Electronic

ISSN: 1558-0687 February 2020.

Guan Gui, Yiwei Song, Hao Huang, Jie Yang, and Hongji

Huang, “Deep learning for super-resolution channel

estimation and DOA estimation based massive MIMO

system”. IEEE Transactions on Vehicular

Technology , Volume: 67, Issue: 9, September 2018.

Hongji Huang, Song Guo, Guan Gui, Zhen Yang, Jianhua

Zhang, Hikmet Sari, and Fumiyuki Adachi, “Deep

INCOFT 2025 - International Conference on Futuristic Technology

912

learning for physical-layer 5g wireless techniques:

Opportunities, challenges and solutions”,

arxiv:1904.09673v, 21 APR 2019.

Hongji Huang, Yiwei Song, Jie Yang, Guan Gui, and

Fumiyuki Adachi, “Deep-learning-based millimeter-

wave massive mimo for hybrid precoding”, IEEE

Transactions on Vehicular Technology, 68(3):3027–

3032, July,2019.

Yuxin Lu; Peng Cheng; Zhuo Chen, Wai Ho

Mow, Yonghui Li, Branka Vucetic, “Deep Multi-Task

Learning for Cooperative NOMA: System Design and

Principles”, IEEE Journal on Selected Areas in

Communications ,Volume: 39, Issue: 1, January 2021.

Performance Analysis of Cooperative Non-Orthogonal Multiple Access System Using Deep Learning Technique

913