One of the most persistent safety gaps in driving safety is blindspots – the hidden pedestrians or objects obscured by large trucks or buildings – that can lead to major accidents. But for autonomous vehicles equipped with state-of-the-art technology, blindspots could become a problem of the past.
UCF researchers have created an award-winning AI-based modeling system that fuses together data from vehicles, roadside cameras and local infrastructure to create an aerial view of the road – essentially enabling the vehicle to see what drivers cannot.
This project placed third in the DriveX Grand Challenge, which is hosted by the Institute of Electrical and Electronics Engineers and the Computer Vision Foundation through the annual Conference on Computer Vision and Pattern Recognition (CVPR). This challenge tasked participants with developing systems capable of detecting and identifying objects in complex urban environments.
Postdoctoral scholar Muhammed Shahbaz ’25 and doctoral student Thamed Chowdhury, from the Department of Civil, Environmental and Construction Engineering, created a cooperative sensing model that integrates visual information from cameras and LiDAR sensors to improve object detection accuracy. By combining data from different sensing technologies, the team developed a model capable of identifying objects that might otherwise remain hidden from a single vehicle’s view.
What gave the Knights an edge over the competition this year was the way that they trained.
“This year, we invested in synthetic data for AI training,” says Associate Professor Shaurya Agarwal, who supervises Shabaz in the Urban Intelligence and Smart City Lab. “Good data is necessary for computer perception.”
While autonomous vehicles have made significant progress in recent years, the researchers acknowledge that navigating unpredictable city environments remains one of the field’s greatest challenges. As cities explore connected vehicle technologies and intelligent transportation networks, advances in cooperative sensing become increasingly essential to improve both safety and efficiency on the roads.
“The idea behind cooperative perception is simple – no single vehicle has a complete view of the world,” Agarwal says. “With vehicle-mounted sensors and LiDAR sensors at intersections, these vehicles can share knowledge of roadside perception with each other, increasing safety for drivers and pedestrians.”
- Written by Marisa Ramiccio