Optimizing Edge Computing Performance through Advanced Pruning Techniques and Multi-Task Learning for Enhanced 5G features Prediction.

• Grant number: 24/18359-0
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: December 01, 2024
• End date: November 30, 2028
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Fabio Luciano Verdi
• Grantee: Mina Kaviani
• Host Institution: Centro de Ciências em Gestão e Tecnologia (CCGT). Universidade Federal de São Carlos (UFSCAR). Campus de Sorocaba. Sorocaba , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação (FEEC)
• Associated research grant: 21/00199-8 - SMART NEtworks and ServiceS for 2030 (SMARTNESS), AP.PCPE

Abstract

The rapid advancement of 5G technology has caused a fundamental shift in communication, driving the imperative for heightened Edge Computing (EC) capabilities to cater to the evolving demands for predictive features. As 5G technology continues to progress, it is essential to enhance EC to meet the increasing requirements for predictive functionalities. One of the goals of this research is to predict 5G indicators such as Channel Quality Indicator (CQI), Throughput, and Latency using Deep Neural Network (DNN) models. Anticipating these metrics is vital for upholding and enhancing the efficacy, dependability, and user satisfaction within 5G networks. Implementing these models in Edge environments can pose significant challenges due to high memory and computing power requirements. To address issues such as memory constraints and processing speed, we plan to employ techniques like Multi-Task Learning (MTL) and model pruning. By applying advanced pruning methods and MTL algorithms, we aim to streamline data processing while enhancing predictive accuracy. Through data pruning, we reduce computational strain on Edge devices, facilitating faster processing. Simultaneously, MTL enables the concurrent prediction of multiple 5G features, ensuring robustness and accuracy. By using these methods, we anticipate not just improved EC efficiency but also a smoother user experience as wireless networks evolve, with substantial enhancements in computational efficiency and prediction accuracy. We intend to evaluate our approaches using publicly available 5G datasets, varying different scenarios such as pedestrians and vehicular movements.