NN-based Puncturing for eMBB-URLLC Coexistence in 5G Wireless Networks.

Show simple item record

dc.contributor.author Ahamed, R.S.
dc.contributor.author Gunarathne, R.M.S.
dc.contributor.author Gunarathne, G.W.L.
dc.contributor.author Ajanthan, A.A.
dc.contributor.author Weerasinghe, T.N.
dc.contributor.author Dhahlan, A.S.A.
dc.date.accessioned 2025-07-07T04:12:00Z
dc.date.available 2025-07-07T04:12:00Z
dc.date.issued 2025-06-04
dc.identifier.citation Ahamed, R. S. , Gunarathne, R. M. S., Gunarathne, G. W. L., Ajanthan, A. A., Weerasinghe, T. N. & Dhahlan, A. S. A. (2025). NN-based Puncturing for eMBB-URLLC Coexistence in 5G Wireless Networks. 22nd Academic Sessions & Vice – Chancellor’s Awards, Faculty of Agriculture, University of Ruhuna, Sri Lanka. 63. en_US
dc.identifier.issn 2362-0412
dc.identifier.uri http://ir.lib.ruh.ac.lk/handle/iruor/19736
dc.description.abstract Ultra-reliable low latency communication (URLLC) and enhanced mobile broadband (eMBB) are two of the main service components in 5G wireless networks. While URLLC demand ultra-high reliability and low latency. eMBB require high data rates and spectral efficiency. The coexistence of eMBB and URLLC in the radio frequency resources is challenging due to the trade-off among latency, reliability, and spectral efficiency. Existing mechanisms, such as dynamic resource allocation, puncturing, network slicing, and interference management, often suffer from complexity, inefficiency, and adverse effects on performance of eMBB users. This project proposes a Neural Networks (NN) assessed puncturing mechanism for coexistence of eMBB and URLLC traffic in the downlink where selected eMBB users’ resources are punctured with URLLC traffic to fulfill the strict latency requirements of URLLC traffic while minimizing the impact on eMBB users. Herein, a fully connected feedforward NN is utilized for multi-output regression tasks, trained on datasets that reflect initial Block Error Rate (iBLER) and Modulation Coding Scheme (MCS) values of eMBB traffic. Both models predict the optimal eMBB users to puncture, minimizing disruption while meeting URLLC latency requirements. To evaluate performance, four NN architectures were tested. They are Baseline, Deep NN, Dropout NN, and Batch Normalization NN. While Deep NN showed superior accuracy, the Baseline NN was selected for its balanced performance and computational efficiency. Balanced performance avoids overfitting or underfitting, ensuring good generalization to unseen data with acceptable accuracy on training and validation datasets. Additionally, K means clustering mainly used to get an acceptable puncturing position while considering both iBLER and MCS. Experimental results demonstrate that this method effectively accommodates URLLC traffic with minimal eMBB degradation, validating its real-time applicability in dynamic 5G environments. This study lays the groundwork for more adaptive and efficient resource allocation strategies and suggests exploring higher numerologies for enhanced flexibility in future research. en_US
dc.language.iso en en_US
dc.publisher Faculty of Agriculture, University of Ruhuna, Sri Lanka. en_US
dc.subject eMBB en_US
dc.subject Neural networks en_US
dc.subject Puncturing en_US
dc.subject Resource Allocation en_US
dc.subject URLLC en_US
dc.title NN-based Puncturing for eMBB-URLLC Coexistence in 5G Wireless Networks. en_US
dc.type Article en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account