NVIDIA is reportedly developing an in-house self-driving robotaxi project, with a potential announcement expected soon after an upcoming company all-hands meeting, according to a report from UDN.
The robotaxi is planned to be based on NVIDIA’s new DRIVE AGX Thor system and will adopt a new, one-step technology route that “only uses an end-to-end neural network.” This approach involves strengthening the neural network’s training through a world model created with simulation technology, a process noted as being similar to the system used for Tesla’s Full Self-Driving technology.
Due to this technical focus, sensor components and lens applications are described as “crucial” for the robotaxi’s development. Taiwanese companies, including Tong Hsing Electronics and Yanguang, are expected to “benefit greatly” from supplying components for the project. The initiative is also supported by OmniVision, the world’s second-largest CMOS image sensor (CIS) supplier, which recently announced its entry into the NVIDIA DRIVE AGX Thor ecosystem. Industry sources believe this partnership will help NVIDIA build its robotaxi, highlighting the importance of image sensors. Tongxin Electronics, OmniVision’s primary packaging and testing partner, is also expected to benefit from the collaboration.
NVIDIA CEO Jensen Huang plans to assign the project to senior director Ruchi Bhargava, as per industry sources. This project would position NVIDIA to compete with companies such as Tesla and Chinese electric vehicle manufacturer BYD in the autonomous driving sector. More details about the in-house self-driving robotaxi platform are anticipated in the weeks and months ahead.
