Autonomous vehicles are designed to move across cities, while robots are built to move through manufacturing plants. The wireless networks these systems rely on need to be able to keep up.
That challenge is at the center of new research by Ye Hu, an assistant professor in the Department of Industrial and Systems Engineering at the University of Miami College of Engineering. Backed by the National Science Foundation, Hu is developing an AI-powered networking system designed to help autonomous machines stay connected as they move.
The work is part of a three-year, $700,000 collaborative project with the University at Buffalo. The University of Miami's share of the award is $233,333, with Hu serving as principal investigator for the University team.
A moving target
The project addresses a fundamental shift expected in next-generation wireless networks. As more devices become connected and wireless systems evolve to accommodate them, engineers are exploring higher-frequency signals and more advanced antenna systems.
"We will leverage higher frequency bands and larger antennas in next-generation wireless networks, so the electromagnetic signal has a different shape," said Hu. "It used to be about steering a beam. Now, we can focus energy on a specific spot in 3D space. And that spot is no longer limited to a few centimeters away—it can be tens of meters away, or even around 100 meters with sufficiently large antenna arrays."
So what happens when the device receiving that signal is constantly moving?
Hu's team is working to redesign how networks respond to that movement. Rather than expecting a device to remain within a predictable wireless connection, the researchers want the network to sense its surroundings and use artificial intelligence to adapt as the device moves and help maintain its connection.
The wireless signal itself can help identify where a device is and whether objects are blocking its connection, essentially creating a map of where communication is likely to be strongest. That information can then feed an AI model that makes decisions about how the network and the autonomous system should respond.
Those decisions can happen on both ends. A network could adjust how it communicates with a moving device, and an autonomous system could also factor connectivity into the route it takes.
"It is no longer just about moving autonomous vehicles over the shortest route," Hu said. "it also about how to keep them connected to the network along the route."
Building the testbed
While the research begins with theory, simulations and computer models, Hu also plans to test those ideas on something that actually moves. She is building an autonomous driving testbed using small AI-enabled race cars equipped with antennas, cameras and sensors.
Starting with one vehicle, Hu and her graduate student, Jiahao Wang, will test whether the networking and AI approaches developed through the project can maintain communication as the car moves.
Hu eventually plans to add more vehicles that can sense and communicate with one another, creating a small-scale autonomous driving network. The miniature network will give researchers a physical test environment where they can see whether ideas developed through models and simulations work on moving autonomous systems.
Bringing research opportunities to students
The testbed will also bring the research into the classroom. Hu teaches Decision Making and Reinforcement Learning, a course for undergraduate and graduate students focused on using AI to make decisions in real-world settings.
As the testbed develops, she plans to give students opportunities to apply concepts they are learning in class.
"The students will have hands-on experience in implementing AI approaches into real-world equipment, not just reading the material or running tests on their computers," Hu said.
