Shuang Li

Topic

Towards Sustainable and Realtime Wireless Body Area Networks: Smart Scheduling and Session-Specific Design with Deep Reinforcement Learning

Department of Electrical and Computer Engineering

Date & location

• Wednesday, August 6, 2025
• 9:00 A.M.
• Virtual Defence

Examining Committee

Supervisory Committee

• Dr. Hong-Chuan Yang, Department of Electrical and Computer Engineering, University of Victoria (Supervisor)
• Dr. Lin Cai, Department of Electrical and Computer Engineering, UVic (Member)
• Dr. Yang Shi, Department of Mechanical Engineering, UVic (Outside Member)

External Examiner

• Dr. Xiaolan Liu, Department of Electrical, Electronic and Mechanical Engineering, University of Bristol

Chair of Oral Examination

• Dr. Adel Guitouni, Gustavson School of Business, UVic

Abstract

Wireless body area networks (WBANs) are essential for health monitoring applications but face ongoing challenges in balancing the trade-off between energy efficiency and information freshness. On one hand, biosensors demand high energy efficiency, as battery charging or replacement can be challenging. On the other hand, status updates from biosensors should be sent to the data sink with high reliability and minimal latency. This thesis presents a series of optimal transmission designs for WBANs, leveraging machine learning (ML) and deep reinforcement learning (DRL) technologies to address the diverse, stringent, and often conflicting requirements of sustainable and real-time healthcare systems.

Across diverse WBAN scenarios, we first investigate energy-efficient transmission designs that minimize energy consumption while satisfying reliability and latency constraints through optimal parameter configuration. Second, we develop highly innovative transmission strategies, minimizing the age of information (AoI), through reliable and low-latency communication with effective scheduling strategies. Third, we propose a multi-objective optimization (MOO) approach to balance trade-offs among conflicting performance metrics.

For each transmission design problem, we formulate the corresponding optimization problem and convert it into a Markov Decision Process (MDP). Based on the characteristics of the state and action spaces, we develop tailored ML/DRL-based solutions to learn near-optimal transmission policies. To address modeling inaccuracies and environmental dynamics, we introduce online tuning methods that adapt the learned policies using real-time experiences.

We further highlight key trade-offs in transmission design and conduct comparative analyses through selected numerical examples. The results demonstrate the effectiveness of the proposed ML/DRL-based solutions in addressing complex transmission designs in WBANs, enabling energy-efficient and freshness-sensitive communication, and advancing the practical deployment of sustainable and real-time health monitoring systems.