Interestana
Home/News/Waymo Reveals Compute 'Brain' Powering Robotaxis
The Verge3 min read

By Interestana AI Editorial — AI-drafted, human-overseen. How we report

Waymo Reveals Compute 'Brain' Powering Robotaxis

Waymo, the autonomous vehicle subsidiary of Alphabet Inc., has publicly disclosed significant details about the sophisticated onboard computer system that serves as the 'brain' for its fleet of robotaxis. This powerful compute unit, housed in the trunk of Waymo's vehicles, is capable of performing up to one quadrillion operations per second, enabling the complex decision-making required for autonomous driving. The revelation marks the first time Waymo has shared specific information about this critical component of its self-driving technology.

The system is built around a custom-designed chip, referred to as the "Waymo System on Chip" (SoC), which is developed in-house. This SoC is central to processing the vast amounts of data collected by the robotaxis' sensors, including cameras, lidar, and radar. The computational power of the Waymo SoC is crucial for tasks such as object detection, path planning, and real-time navigation, allowing the vehicle to perceive its environment and react safely to dynamic traffic conditions. Waymo has indicated that the development of this custom silicon is a key differentiator, providing them with greater control over performance and efficiency.

In addition to the custom SoC, the 'brain' incorporates other high-performance computing elements. Waymo has detailed the use of NVIDIA GPUs, which are known for their parallel processing capabilities and are widely used in AI and machine learning applications. These GPUs work in conjunction with the Waymo SoC to accelerate the complex neural network models that underpin the autonomous driving software. The integration of these powerful components allows the robotaxi to process sensor data and make driving decisions with a level of speed and accuracy comparable to, or exceeding, human drivers.

The company also highlighted the redundancy and safety features integrated into the system. Waymo's autonomous driving system is designed with multiple layers of redundancy to ensure that if one component fails, another can take over, maintaining safe operation. This includes backup power systems and diverse sensing modalities. The compute 'brain' is not only responsible for the primary driving functions but also for monitoring the vehicle's health and performance, ensuring the overall safety and reliability of the Waymo ride-hailing service. Waymo's commitment to developing its own hardware and software stack underscores its strategy to maintain a competitive edge in the rapidly evolving autonomous vehicle industry, aiming for scalable and robust deployment of its technology.

Original source — read the full reporting at the publisher:

Read on The Verge

Get the weekly AI digest

AI news + new model releases, weekly. Drafted by our agents, reviewed by humans.

Read next