Research Seminar: Harnessing Real-Time Machine Learning Based Prediction Model in Advancing Wireless Communications and Networks towards the Next Generation
Title: Harnessing Real-Time Machine Learning Based Prediction Model in Advancing Wireless Communications and Networks towards the Next Generation
Date: 23 May 2025 (Friday)
Time: 1030am
Venue: EEE Executive Seminar Rm (S2.2-B2-53)
Abstract:
In recent years, researchers have shown an increasing interest in utilizing AI for wireless communication systems and networks. Various machine learning (ML) algorithms, including deep neural networks (DNN), reinforcement learning (RL), long short-term memory (LSTM), and LLM have been applied to all layers of the wireless protocols to improve system performance and enrich user experience quality (QoE) by facilitating adaptation to variations and uncertainties in electromagnetic environments, mobility, user devices, and network traffic and loads. Next-generation wireless networks are currently in high demand globally, recognized as a critical infrastructure for nations. Presently, next generation wireless networks face numerous demanding criteria, including high data rates reaching Gbps, extensive seamless coverage, simultaneous large-scale access, and highly dependable low-latency links. The complex conditions of channels, noise, interference, as well as circuit and module limitations, render it impractical to account for all variables in the design of wireless communication systems and networks. Moreover, while transmitting a packet or a cluster ranging from several to thousands of bits, the temporal fluctuations in channel traits could greatly impede its successful delivery. The time-varying fluctuations of channels, network traffic, and QoE in next generation wireless services pose significant challenges for traditional communication methods, prompting the consideration of advanced Machine Learning (ML) or Artificial Intelligence (AI) techniques as promising solutions.
The statistical-model based approaches as well as the newly emerging data-driven DNN models highlighted in recent literature are deemed unsuitable for next generation wireless networks according to current assessments. In general, the two methods involve extracting main features from data gathered across various scenarios. The communication system tailored to suit most training scenarios statistically, yielding moderate improvements, though not ideal across all situations. The established integration of a neural network for machine learning follows two distinct stages: training initially and then inferring continually. To this end, a novel real-time machine learning (RTML) concept is proposed in our recent effort, transcending the limitations between training and inference, enabling the seamless execution of sensing, learning, and prediction tasks iteratively in real-time, say in millisecond scale. The AI empowered transceiver will be facilitated to categorize and model the environments and therefore the transceiver will “evolve” to fit well the new applications and missions in the future.
In this presentation, we emphasize some of our recent research and findings in these new areas.
Speaker Bio:
Weidong Xiang is a professor in the Electrical and Computer Engineering Department at the University of Michigan-Dearborn. He founded and directed the Center for Vehicular Communications and Network Laboratory at UMD, specializing in machine learning applications in wireless communications, the Internet of Things, vehicular networks, and autonomous robotics. He has authored over 100 technical papers for international journals and conferences. As the principal investigator, Dr. Xiang has secured more than 20 grants totaling over $2,000,000 in recent decades. His funding sources include the NSF (4 grants), NSA(1), DoE (1), Ford (2), GM (1), LGE (1), CISCO (1), Mcity (1), and the University of Michigan (6). He has also served as an Associate Editor or Editor for publications such as IEEE Communications Magazine, EURASIP Wireless Communications Systems, SAE Journals, among others.
Date and Time
Location
Hosts
Registration
- Date: 23 May 2025
- Time: 02:30 AM UTC to 03:30 AM UTC
-
Add Event to Calendar
- 50 Nanyang Avenue
- Singapore, Singapore
- Singapore 639798
- Building: NTU EEE
- Room Number: EEE Executive Seminar Rm (S2.2-B2-53)