Associate professor in ECE earns $300K NSF award to help intelligent systems learn better from one another
Published: Sep 1, 2026 7:35 AM
By Joe McAdory
Xiaowen Gong was recently awarded $300,000 by the National Science Foundation for his three-year study, “Federated Reinforcement Learning: From Heterogeneous Environments to Offline Data.”
How can autonomous vehicles, robots and other intelligent systems learn from one another and make better decisions collectively when they operate in different environments? That's what Xiaowen Gong will soon find out.
Gong, the Godbold Associate Professor in the Department of Electrical and Computer Engineering who specializes in wireless engineering, was recently awarded $300,000 by the National Science Foundation (NSF) for his three-year study, “Federated Reinforcement Learning: From Heterogeneous Environments to Offline Data.”
His project focuses on federated reinforcement learning, an artificial intelligence (AI) tool that allows multiple intelligent agents to learn collaboratively by sharing knowledge gained through individual experiences.
“Compared to conventional reinforcement learning without any collaboration, agents can share their learning experiences in an intelligent way so that they all can improve their learning performance,” Gong said. “Different agents will exchange useful information or knowledge learned from their individual interaction with the environment, and then they try to improve their decision-making based on other agents' history of interacting with the environment.”
Helping intelligent systems learn collectively presents new challenges as they become increasingly connected, whether they are autonomous vehicles navigating different roads or robots working together in warehouses.
The first challenge, Gong said, was heterogeneous environments, where intelligent agents experience different operating conditions. The second challenge, offline data, is where systems learn using limited historical data instead of collecting new information.
“Different vehicles have different technical performance, and they may drive in different environments, for example, different vehicle models, different roads, or different weather,” Gong said. “All of these elements are part of the environment for this learning problem, and they can be very different for different vehicles.
“We might have only some past data which has already been obtained from the history of a vehicle, but we may not be able to let it drive itself in the environment we want. In this case, we have limited data and we cannot obtain new data. How can we make the best use of this offline data to improve performance?”
Gong said research outcomes of this project have the potential to enable intelligent control and management of wireless and computer networks, and also support various emerging AI applications over networked systems, such as collaborative robotics, connected and autonomous vehicles, multi-user mixed reality.
“Many applications can be impacted by this research, but I think autonomous driving is the most important,” Gong said. “As autonomous driving gets better, more people use it in their real lives. It's making an impact in the real world, and hopefully our research can improve that application.”
Media Contact: , jem0040@auburn.edu, 334.844.3447