ECE Team Wins Best Student Paper Award at 2025 WiOpt Conference
In joint work with Professor Igor Kadota, Yubo Zhang and Pedro Botelho developed a machine learning-based dynamic spectrum access algorithm for decentralized wireless networks
Northwestern Engineering’s Yubo Zhang and Pedro Botelho earned the Best Student Paper Award at the 23rd International Symposium on Modeling and Optimization in Mobile, Ad hoc, and Wireless Networks (WiOpt) last month in Linköping, Sweden.
Zhang and Botelho are members of the Communications and Networking (Commnet) Laboratory. Zhang is graduating this month with a master’s degree in electrical engineering. Botelho is a visiting student who earned a bachelor’s degree in electrical engineering from the Aeronautics Institute of Technology in Brazil.
The annual WiOpt conference showcases state-of-the-art, theoretical, experimental, and empirical research related to the modeling, performance evaluation, and optimization of networks.
“I’m deeply grateful to the WiOpt 2025 committee for this recognition, which affirms the effort and passion I’ve poured into this work,” Zhang said.
The winning paper, titled “Fair Dynamic Spectrum Access via Fully Decentralized Multi-Agent Reinforcement Learning,” is both Zhang and Botelho’s first published academic research paper. The work was coauthored by adviser Igor Kadota, assistant professor of electrical and computer engineering at the McCormick School of Engineering, and Trevor Gordon and Gil Zussman (Columbia University).
In the paper, the team examined machine learning-based mechanisms to effectively manage limited spectrum resources in decentralized wireless networks. They considered whether devices can learn to transmit autonomously without any form of coordination or cooperation, unlike protocols in traditional wireless networks which direct how and when devices transmit data.
The team developed a fairness-driven algorithm for decentralized communication networks in which Fair Share Reinforcement Learning (FSRL) agents learn and continuously improve their transmission strategy.
“Unlike existing machine learning solutions that rely on information exchange among devices, the FSRL agents learn to transmit efficiently and fairly with distributed training,” Zhang said.
Using numerical simulations, the team also demonstrated that FSRL agents can successfully overcome harmful interference caused by jammers maliciously occupying a selected band of frequency spectrum to block communication.
“Our FSRL solution is a first step toward enabling devices in a decentralized wireless network to learn to transmit by themselves while adapting to time-varying network conditions,” Kadota said.
While traditional wireless infrastructure, such as 5G and Wi-Fi, relies on centralized network providers, decentralized wireless networks are emerging as connectivity solutions for applications, including vehicle networks where cars transmit position and velocity information to improve road safety. Decentralized wireless networks also allow networks of robots to transmit status information, enabling search and rescue and other collaborative tasks.
As a next step in the research, the team is implementing FSRL in a testbed with several programmable radios to evaluate its performance in real-world wireless environments. In parallel, the researchers are also improving FSRL to further reduce its convergence time, enabling faster adaptation to dynamic conditions.
“Receiving the Best Student Paper Award for my first paper after a lot of hard work with Yubo Zhang and Professor Kadota is a great honor,” Botelho said. “This prize reinforces my dream of applying for and pursuing a PhD program to conduct research and share new knowledge.”