RideFlux Achieves Top Performance in CVPR 2025 Autonomous Driving Challenge

Autonomous driving software startup RideFlux has made a name for itself on the global stage with its innovative technology. On the 11th, it proudly secured 3rd place in the vision-based End-to-End (E2E) autonomous driving category of the 'Waymo Challenge,' held at the CVPR 2025 AI conference's autonomous driving workshop. This is the fruit of successful industry-academia collaboration with Hanyang University's IRCV research team and a significant achievement that once again demonstrates the potential of domestic autonomous driving technology. In this challenge, RideFlux, in collaboration with the Hanyang University research team, proposed a model that effectively plans the movement paths of autonomous vehicles based on the vast camera video data released by Waymo. Their 'Swin-Trajectory' method is an innovative algorithm that precisely analyzes the vehicle's past history and image context using AI technology to predict future trajectories. It boasts not only excellent accuracy but also overwhelming computational speed, executing in approximately 14ms on a commercial GPU, earning it recognition as a solution optimized for autonomous driving environments where real-time processing is essential. Notably, the vision-based E2E autonomous driving category saw participation from 29 leading R&D teams worldwide, including NVIDIA, Xiaomi, EPFL (Swiss Federal Institute of Technology Lausanne), and Mila AI Institute (Canada), making it the most intensely competitive section among all categories of the Waymo Challenge. Securing 3rd place amidst such global competition is a valuable achievement that proves RideFlux's technology is among the world's best. E2E autonomous driving is currently the most notable global technology trend in the industry. This technology, designed for AI models to learn from various driving video data and integrally perform all autonomous driving processes such as perception, prediction, judgment, and control, is emerging as a core of future mobility. Professor Hwang Soon-min of Hanyang University's Department of Automotive Engineering commented, "This is an exemplary case of industry-academia cooperation achieved through the dedicated efforts of students and close collaboration with RideFlux," expressing his pleasure in being able to showcase the potential of domestic technology to the world. Park Jung-hee, CEO of RideFlux, also emphasized, "It was an opportunity to reconfirm the infinite potential of End-to-End autonomous driving and the competitiveness of purely domestic technology," revealing his ambition to "continue strengthening R&D capabilities based on high-quality data to complete the most reliable autonomous driving service." This award will be a significant milestone, showing that RideFlux has taken another step closer to the 'safest and most efficient future of autonomous driving' it envisions.