Hello, welcome to Zhouyou Gu’s website!

Biography

(M,IEEE) I received the B.E. (Hons.) and M.Phil. degrees from The University of Sydney (USYD), Australia, in 2016 and 2019, respectively, and the Ph.D. degree from Centre for IoT and Telecommunications at the School of Electrical and Information Engineering, USYD, in 2023.

Throughout my studies, I was fortunate to work under the supervision of Dr. Wibowo Hardjawana and Prof. Branka Vucetic, which provided me with invaluable guidance in my research endeavors. My Ph.D. thesis focuses on “Scheduler Designs in Wireless Networks” and my works are published in prestigious journals and conferences, e.g., TON, JSAC and TWC etc.

I was a research assistant of Prof. Branka Vucetic at USYD in 2023 and am currently a causal research fellow of Prof. Jinho Choi at Deakin University and will be joining Singapore University of Technology and Design (SUTD) soon as a research fellow of Prof. Jihong Park.

My research interests span in graph theory and machine learning in wireless networks.

🤝Collaborations from both industry and academia are highly welcomed. If you are interested in my researches, feel free to reach out!

Research Outputs

My research has contributed to the advancement of wireless network technologies, which includes but not limited to the following areas.

Algorithms

Algorithms have been designed in my research to address challenges in different aspects of networks.

• AC-GRL enables flexible network-wise coordination by embedding neural networks in graph theory methods. Paper
  🎯To the best of our knowledge, this is the first work to train neural networks to flexibly construct the optimal graph representing the impact of interference on specific network performance.

• MVWO accelerates convergence of scheduling policy using statistical information, e.g., mean and variance of network states. Paper
  🎯To the best of our knowledge, this is the first work to accelerate max-weight schedulers’ convergence using limited prior knowledge of statistical channel state information, e.g., the mean and variance of users’ channel states, that costs only few samples to estimate.

• K-DDPG improves quality of service of reinforcement learning in networks by integrating domain-expert knowledge. Paper
  🎯To the best of our knowledge, this is the first work to provide an end-to-end solution of a deep-reinforcement-learning-based scheduler design in cellular networks from theoretical formulation to a real-time prototype.

Architectures

My research also involves network system architecture designs.

• ac-grl-wi-fi, a gym-like simulation platform for network-wise time-slot allocation algorithms in Wi-Fi, prototyped using NS-3. Github
  🎯To the best of our knowledge, this is the first gym-like simulation platform for machine-learning-based slot-allocation in Wi-Fi.

• drl-5g-scheduler, an online neural network fine-tuning archiecture for 5G radio access networks (RANs), operating at per millisecond level, prototyped using srsRAN, USRP Software Defined Radio (SDR) and Nvidia GPU. Github
  🎯To the best of our knowledge, this is the first implementation to use a neural network to schedule transmissions per transmission time interval in a real-world RAN and simultaneously fine-tune the neural network’s weights online.

• OPSCH, a vendor-neutral programming interface for schedulers in 5G networks, prototyped using OpenAirInterface and srsRAN. Paper
  🎯To the best of our knowledge, this is the first programming interface that can online update the scheduler logics in RANs without interrupting the services.

• M-AP, a multi-tenant virtualization architecture with channel and beacon controls in Wi-Fi, prototyped using WARP SDR Platform. Paper
  🎯To the best of our knowledge, this is the implementation considering interference information for joint overlapping channel and service set identifier allocation in a multi-tenant Wi-Fi networks.

Collaborations

It is my honor to have had collaborations with colleagues from

• Morse Micro (A fabless semiconductor startup): Together, we investigated the interference management using restricted access window (RAW) in Wi-Fi Halow networks. (2022-2023)

• SmartSatCRC (Australia’s leading space research center): Together, we investigated advanced satellite communications for high rate and dynamic service delivery. (2020)

• Telstra (Australia’s largest telecommunications company): Together, we investigated open programmable schedulers of 4G/5G networks and open radio access networks (O-RANs). (2018-2019)

Awards

I have proudly collected national-wise and university-wise scholarships, including

• the Research Training Program Stipend, 2019
• the Postgraduate Research Supplementary Scholarship in Cellular IoT Networks, 2019
• the Postgraduate Research Supplementary Scholarship in Online Machine Learning for Next Generation Autonomous IoT Networks, 2020
• the Postgraduate Research Supplementary Scholarship in Development of AI Schedulers for Satellite Networks, 2021
• the Postgraduate Completion Scholarship in Development of Resource Allocation Techniques for Long Range IEEE 802.11ah System, 2022