International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 10 | Oct 2025
p-ISSN: 2395-0072
www.irjet.net
Survey on AI based Network Anomaly Detection for next generation wireless communication system for performance enhancement. Amithesh Y1, Dhyuthi K A2, Rathna V3, Priyadarshini K Desai4 1Department of Electronics and Communication Engineering, BNMIT, Bangalore, India 2Department of Electronics and Communication Engineering, BNMIT, Bangalore, India 3
Department of Electronics and Communication Engineering, BNMIT, Bangalore, India
4 Associate Professor, Department of Electronics and Communication Engineering, BNMIT, Bangalore, India
---------------------------------------------------------------------***--------------------------------------------------------------------and propagation conditions, while an end-to-end voice Abstract - An AI-driven channel modeling framework in transmission pipeline validates real-time feasibility and reliability, highlighting gains in accuracy, adaptability, and scalability over traditional models for next-generation wireless design and testing.
MATLAB evaluates 4G, 5G, and 6G links across terrains, elevations, and obstacle densities using an interactive GUI for BER–SNR, BER–distance, and SNR–distance analyses. A neural network trained on synthetic Rayleigh fading predicts effective channel gain and path loss with environment-aware corrections, improving accuracy and adaptability over analytical baselines. An end-to-end voice transmission through the learned channel validates practical reliability and latency awareness, indicating benefits for robust, next-generation wireless design and evaluation.
2. LITERATURE SURVEY [1]
Millimeter Wave Channel Modeling via Generative Neural Networks: Demonstrates that GAN-based synthesis can reproduce realistic mmWave spatial– temporal multipath statistics, aiding data augmentation and beam management evaluation beyond fixed parametric models.
[2]
Attention-Guided Wireless Channel Modeling for Next-Generation Networks: Uses attention to focus on dominant paths and blockage, improving generalization across environments and reducing prediction error for channel reconstruction and prediction tasks.
[3]
Path Loss Modeling Based on Neural Networks for Diverse Terrains: Shows terrain-aware neural and ensemble models significantly lower RMSE versus logdistance baselines, enabling environment-conditioned planning across urban, suburban, and rural settings.
[4]
Machine Learning in 6G Wireless Communications: Surveys model-driven and data-driven methods for channel estimation, link adaptation, and resource allocation, highlighting robustness, sample efficiency, and interpretability as core challenges for 6G adoption.
[5]
Advances and Future Challenges on 6G Wireless Channel Measurements and Models: Identifies gaps in multi-band, mobility-aware datasets and calls for physicsinformed learning to capture sparsity, blockage dynamics, and THz propagation nuances.
[6]
Adaptive/Implicit Deep Learning for Channel Tasks: Indicates implicit neural representations can capture finegrained channel variations with fewer parameters and continuous function priors, improving fast adaptation and accuracy with limited pilots.
Key Words: AI-based channel modeling; 4G/5G/6G
wireless systems; millimeter-wave and sub-THz propagation; path loss prediction; Rayleigh and Rician fading; BER–SNR analysis; received signal strength (RSSI); OFDM and link adaptation; QPSK, 16-QAM, 64-QAM, 256-QAM; environment-aware estimation (terrain, elevation, obstacle density); neural networks for channel gain and path loss; MATLAB simulation GUI; voice transmission over learned channel; physics-informed learning; dataset curation and anomaly filtering for robust training.
1. INTRODUCTION The evolution from 4G to 5G and towards 6G has amplified demands for high data rates, ultra-low latency, and robust connectivity across diverse, dynamic environments, making accurate channel modeling central to reliable system design and evaluation. Classical analytical and statistical models (e.g., Rayleigh, Rician, log-distance path loss) provide useful baselines but often fall short in capturing nonlinear, time-varying behaviors driven by mobility, blockage, multipath richness, terrain, elevation, and obstacle density, leading to gaps in predicting key metrics such as BER, SNR, and RSSI across heterogeneous scenarios. AI-driven approaches address these challenges by learning effective channel gain, path loss, and fading dynamics from data, enabling adaptive, environment-aware prediction and faster performance assessment across 4G, 5G, and 6G settings. Within a MATLAB-based workflow, an interactive interface supports BER–SNR, BER–distance, and SNR– distance analyses under selectable standards, modulations,
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 246