“Over 20,000 smart driving cars integrated with our DeepRoute IO are driving on public roads, and we expect this figure will grow tenfold next year. In the AI-driven era, we take pride in being a key force in bringing cutting-edge technology and next-level driving experiences to the public.” said Xuan Liu, Vice President at DeepRoute.ai, at the Guangzhou Auto Industry Development Summit, themed “Unleashing new-quality productivity, Opening new horizons”.
DeepRoute IO, first launched in April this year at Beijing Auto Show, has been successfully integrated into over 20,000 mass produced passenger cars, now actively used by consumers as part of their daily driving routine. This week at Auto Guangzhou, DeepRoute.ai showcased its latest beta version, featuring several updates for public test drive.
Guangzhou is a vibrant metropolis known for its bustling streets filled with new energy vehicles and zipping e-scooters, all navigating a maze of narrow alleys and wide boulevards. Amid this dynamic urban landscape, smart driving car integrated with DeepRoute IO effortlessly navigates through crowded streets, skillfully avoiding obstacles like e-scooters and pedestrians, offering a level of precision and safety that ensures a comfortable driving experience in even the most chaotic of cityscapes.
The latest beta version of DeepRoute IO introduces several major updates including:
New features include the recognition and rendering of emergency vehicles such as ambulances, police cars and police officers, providing a comprehensive overview of the surroundings.
U-turn function. The smart driving car smoothly and efficiently handles U-turns with a success rate of over 95%, safely navigating alongside other traffic users, much like an experienced driver.
Whether driving through heavy traffic or encountering lane changes, the success rate for selecting the correct lane has increased to 96%.
Since the launch of the first series model at the end of August, the penetration rate of smart driving technology among consumers has reached 67% by October. With the second series model released in October and the third set to launch by the end of November, DeepRoute is rapidly advancing in collaboration with automakers to mass-produce smart driving cars, with over 10 series models slated for release next year.
Looking ahead, the company is developing DeepRoute IO 2.0, based on Vision-Language-Action (VLA) Model utilizing NVIDIA DRIVE Thor SoC and end-to-end model, with deployment planned for next year. The VLA model combines vision, language and action together, enabling the smart driving system to not only recognize and describe the road, environment, traffic signs, and road participants, but also understand complex scenarios such as negotiation and hidden semantic information, with advanced logical reasoning capabilities. Even in rare or unexpected scenarios, the system will maintain excellent performance thanks to its robust generalization capability.
