Santa Clara, Calif. – has announced plans to develop end-to-end smart driving solutions powered by the centralized compute architecture of the upcoming NVIDIA DRIVE Thor platform, based on the new Blackwell GPU architecture designed for transformer and generative AI workloads. This architecture will be first adopted by passenger vehicles integrated with's smart driving solution, and scheduled for launch next year. began conducting end-to-end (training to deployment) architecture road tests on NVIDIA DRIVE Orin, the predecessor to DRIVE Thor, back in August 2023, validating the viability of the model through significant testing and simulation. DeepRoute will be able to leverage the benefits of their software investment in DRIVE Orin, and take advantage of the scalable DRIVE architecture to easily transition current development efforts to DRIVE Thor once ready.

During his session at GTC, CEO Maxwell Zhou, showcased the videos from these road tests, which demonstrated proficiency in navigating complex scenarios with oncoming vehicles and pedestrians in a natural, polite manner while ensuring safe and efficient driving. Another example showed the vehicle carefully changing lanes to pass when recognizing a taxi dropping off passengers. As the model continues to obtain a wealth of data, it is poised to iterate results and performance that offer a safe, smooth and efficient driving experience.

Unlike traditional autonomous driving solutions that separate into perception, prediction, planning and control modules etc., the end-to-end model integrates all of these modules into one neural network, allowing for no information loss. The model also learns from a vast repository of video clips that fosters insight into more nuanced driving behaviors. One example of this sophistication is the model's capacity to understand courtesy and consideration toward pedestrians during heavy rain, adjusting its actions to take cues from the pedestrian’s behavior.

"The end-to-end model is revolutionary for autonomous driving, signaling the end of 'rule-based' AI and the dawn of a new era that is ‘learning-based’." said Maxwell Zhou, during his GTC session. "High performance SoCs like NVIDIA DRIVE Thor with 1,000 teraflops of performance, are crucial for future success of the end-to-end model. With this technology, we will achieve our goal of artificial general intelligence in robots by integrating smart driving cars with end-to-end model."

"DeepRoute has demonstrated remarkable technology innovation prowess in recent years, evolving into a staple within in the transportation industry for AI driving systems." said Rishi Dhall, Vice President of Automotive, NVIDIA.