Waymo is providing a detailed look at the sophisticated computing systems within its driverless vehicles, emphasizing redundant hardware and backup mechanisms to ensure safety in the event of a primary computer failure. The company has designed its system to make critical driving decisions in milliseconds without relying on external cloud connections or remote human intervention.
Onboard Computing and Redundancy
Waymo's latest computing system operates with a dual-engine approach. Both systems can execute full driving tasks concurrently during normal operation. If one system encounters a fault, the other is designed to seamlessly take over. This internal redundancy means the vehicle does not have to wait for assistance from miles away. The company states that a secondary onboard computer can bring the vehicle to a safe stop if a failure is detected in the primary system. Complementing this are backup braking and steering hardware, with separate power sources dedicated to critical driving systems.
The immediate priority in the event of a significant failure is to maintain vehicle control and achieve a safe state, which may involve stopping rather than completing the journey. This layered approach to system backups is intended to provide a sense of security for passengers riding without a human driver.
Processing Environmental Data
A Waymo vehicle processes vast amounts of data from its sensors, including cameras, lidar, and radar. The latest sixth-generation system employs 13 cameras, alongside lidar and radar, to create a comprehensive understanding of the vehicle's surroundings. This information is fed into the onboard system, which must rapidly interpret it to navigate, steer, and react to potential hazards. Waymo emphasizes that these driving decisions are made entirely onboard, with no need to transmit all data to a central server before acting.
Over the past eight years, Waymo has significantly increased its raw computing power, by approximately 20 times, to reduce the reaction time between perceiving an object and responding to it.
Custom Chip for Sensor Data
Waymo has also developed a custom 5-nanometer application-specific integrated circuit (ASIC). This specialized chip is designed to efficiently process the raw data flowing from the vehicle's sensors, including camera, lidar, and radar information, before it reaches other driving system components. According to Waymo, this ASIC delivers over 1,000 TOPS (trillions of operations per second) of machine-learning performance. This capability allows the system to process data from multiple high-resolution cameras simultaneously and enhances its ability to detect objects in low-light conditions.
Backup Hardware for Critical Functions
Beyond the redundant computing systems, Waymo has integrated backup hardware for essential functions. A secondary braking system is in place to bring the vehicle to a safe stop. Similarly, the steering system features redundant hardware, supported by separate controllers and power supplies for its secondary system. Independent power sources are also critical driving systems to prevent a single power failure from disabling essential operations.
These layers of redundancy have been a focus for Waymo for years, with the company conducting tests for hardware faults, power issues, and software failures. The vehicles are designed to continuously monitor their own systems while operating.
Remote Assistance vs. Remote Control
Waymo clarifies that while remote assistance is available, it does not involve an employee remotely driving the vehicle. Instead, remote agents provide additional context or information if the automated driving system encounters an unusual situation. The vehicle's automated system retains full control and decides whether to utilize the information provided by remote agents. Remote employees do not continuously monitor every vehicle.
Limitations and Ongoing Refinement
Despite these redundancies, Waymo acknowledges that backup computers do not prevent every potential issue. Software can still struggle to interpret novel road situations. The company recently issued a software recall for 3,871 vehicles after they entered closed freeway construction zones. This incident highlighted challenges in how the automated driving system handles changing road conditions, distinct from hardware failures.
Waymo's safety analysis, covering over 220 million fully autonomous miles through March 2026, indicates fewer serious or fatal injury crashes compared to human drivers in comparable areas. However, the company continues to refine its perception, decision-making, and software systems as its fleet expands.
The engineering challenge for a fully autonomous vehicle like Waymo is to handle the complete driving task within its operational domains, often with no human capable of intervening. This requires significant onboard computing power and hardware resilient to the harsh conditions within a vehicle. Waymo's computing hardware utilizes liquid cooling to maintain performance in extreme temperatures.
Rider Preparedness
Waymo advises passengers to utilize the in-vehicle passenger screen or the Waymo app to contact Rider Support if assistance is needed. If a ride needs to end early, the app offers a "Pull over" option, after which the vehicle will find a safe location to stop. In case of unexpected stops, Waymo recommends passengers remain buckled and inside the vehicle unless there is an urgent safety reason to exit. Passengers are instructed not to touch the driving controls. In situations where personal safety is threatened, calling 911 is advised, or using the emergency option within the Waymo app.