Computer Science and Engineering, UT Arlington

What happens when a robot encounters something it has never seen before?

Usually, it freezes up or fails. But new research from the Robotic Vision Laboratory is changing that. Doctoral student Zongyao Lyu is leading this effort alongside William J. Beksi, an associate professor of computer science and engineering.

Their new paper titled "Energy-Based Open-Set Active Learning for Object Classification" was accepted to the International Conference on Pattern Recognition, or ICPR 2026. This breakthrough framework essentially gives artificial intelligence a smart way to handle the unknown.

Right now, when an AI encounters a strange object, it gets confused and requires thousands of new images to retrain. This new system allows the AI to instantly flag something unfamiliar. It then smartly selects only the most important data for a human to label, saving massive amounts of time and computing power. Lyu was also awarded the prestigious UTA Dissertation Fellowship for his research.

This work has immense real-world implications because Lyu is already applying this framework to robotics. Imagine a delivery robot navigating a crowded sidewalk or a warehouse robot handling new products. Instead of freezing up when faced with a strange object, the robot can safely understand its changing environment and adapt on the fly. This brings us closer to truly independent, reliable automation in manufacturing, logistics and daily life.

Code the future. Engineer the impossible.

Our students are developing real-world technology in the lab and driving infrastructure solutions for global tech leaders! 🚀

For our next Student Spotlight, we are featuring Computer Science student Joel Beauregard, who recently wrapped up a successful Software Engineering Internship at Oracle, working on their Private Cloud Appliance (PCA) platform! 💻🌐

Focusing on systems, networking, and infrastructure-oriented engineering, Joel made major contributions to the platform during his time with Oracle. He independently researched and built a customer-requested Network Traffic Analytics feature, analyzing traffic flows to help shape its architecture. He also migrated over 180 unit tests into a new MySQL-backed testing framework, refactoring half of them to improve error handling and boost test performance by roughly 80%.

Outside of his industry work, Joel continues to push his engineering boundaries by building complex systems-level projects, including concurrent TCP port scanners and real-time sensor data pipelines. He plans to continue his impactful work with Oracle while pursuing his degree here at UTA.

We are incredibly proud to see Joel applying high-level systems engineering to major cloud platforms. We cannot wait to see what he builds next!