Rashid Ali, University of Pompeu Fabra Barcelona

Rashid Ali, PhD

• Education

Dr. Rashid Ali (M'20) received his Bachelors in information technology from Gomal University, Pakistan, in 2007.
His Master's degree in computer science under the supervision of Dr. Stanislav Belenki in 2010.
His Master's degree in Informatics under the supervision of Dr. Maria Spante in 2012–2013 from the University West, Sweden.
He received his Ph.D. in Information and Communication Engineering from the Department of Information and Communication Engineering, Yeungnam University, Korea in February 2019.

• Experience

Between 2007 and 2009, he worked for Wateen Telecom Pvt. Ltd. Pakistan as an Executive WiMAX engineer in the Operations and Research Department.
From July 2013 to June 2014, he worked for COMSATS University, Pakistan, as a lecturer. He has more than 10 years of experience in research, academia, and industry in the field of Information and Communication Engineering, and Computer Science. Later, he served for one year as a postdoc research fellow at the Yeungnam University and Sejong University, Korea. He also served for two years as an Assistant Professor at the School of Intelligent Mechatronics Engineering, Sejong University, Seoul, Korea. His research interests include Internet of Things (IoT), 5G, Future WLANs, wireless sensor networks, transport/network/MAC layer protocols, Information-Centric Networking (ICN), Embedded systems/Smart systems, Network-on-a-Chip, Blockchain Networks, Machine/Deep/Reinforcement Learning, and Federated Reinforcement Learning for wireless networks.

He has experience of several national and international research projects as a leading or collaborating researcher. Rashid has three registered patents at the Korea Intellectual Property Office. He authored and co-authored several book chapters related to his research domain. His research contributions include more than seventy research publications in well reputed SCIE/Scopus journals and International Conferences. His other qualifications and certifications include Cisco Certifications, Juniper Networks Certifications, and RedHat Linux training. Rashid also has experience of supervising Ph.D. and Master students at their final project level. He has given talks related to his current research topics at different universities in Korea. He also has experience of chairing conference sessions, as well as organizing ad being a member and editor of a journal.

Recently, a newly launched sixth generation of Wi-Fi (Wi-Fi 6) has shown several breakthroughs, including novel techniques for sharing spectrum resources, such as spatial reuse, multi-user spectrum access, power-saving advances, and so on, which make this standard a very significant step forward concerning its predecessor Wi-Fi 5. While researchers rivet their eyes on Wi-Fi 6, in the IEEE 802.11 Working Group's (WG) bowels, the next generation Wi-Fi, Wi-Fi 7 (IEEE 802.11be), is being developed. At first sight, the upcoming Wi-Fi 7 is nothing but scaled Wi-Fi 6 with the increased number of spatial transmission streams to support extremely high throughput (EHT) and low latency. Although Wi-Fi 7 claims to increase the actual throughput by thirty times in a dense user environment by introducing several new features, questions may arise.Do we know each new feature's individual use? How can we combine these features to achieve higher network performance? Are these features based on machine learning-aware (ML) architectures? Hence, the challenge of optimization of the performance of potential features of Wi-Fi is still open.

At this point, lots of expectations have been directed towards ML as a key enabler of the next generation Wi-Fi 7 and beyond (Wi-Fi 8). Unfortunately, current Wi-Fi 6 and upcoming Wi-Fi 7 are not yet prepared to support the subsequent requirements and usage of ML-based applications. Recently, Reinforcement Learning (RL), one of the ML techniques, has given rise to prominent behaviorist learning techniques for sharing spectrum resources in densely deployed networks. In an RL-enabled spectrum resources sharing technique, a Wi-Fi device optimizes its performance based on its observation from its environment. However, a densely deployed Wi-Fi environment with massively connected networks and devices is of a distributed and dynamic nature, which changes more often.

Thus, relying on the individual local learning models (LLM) like RL leads to higher error variance, especially when spectrum resources are shared among a massive number of connected networks and devices. Therefore, in this project, we propose an ML-aware architecture for Wi-Fi 7 and beyond networks, which uses the Federated Reinforcement learning (FRL) model in Wi-Fi networks for optimized spectrum resource sharing (acronym as WiFi-FRL).