Just as the autonomous driving field is still debating whether "pure vision" or "multi-modal sensing" will lead to the future, Alphabet's Waymo officially announced that its...Sixth-generation autonomous driving system The 6th-generation Waymo Driver has officially begun fully autonomous driving operations and is being applied in...Zeekr's self-driving vehicle "Ojai"And the modern IONIQ 5.
This is not just an iteration of technical specifications, but a crucial turning point for Waymo as it prepares to enter the "10,000-vehicle production capacity" stage. Below is a summary of the core highlights of this update, showing just how powerful this new system, known as the "expansion engine," truly is.
The philosophy of "subtraction": Halving the number of sensors doubles the performance.
The most criticized aspect of previous autonomous driving systems was the expensive and heavy "sensor tower" on the roof. However, in the sixth-generation system, Waymo demonstrated extremely strong hardware and software integration capabilities:
• Visual system:It uses a self-developed 1700-megapixel high-resolution sensor. Because the image quality is good enough, the field of view that originally required more than 10 cameras can now be reduced by more than half, and the dynamic range performance is better in strong light, heavy rain or late at night.
• Processing efficiency:By transferring complex computing power to Waymo's self-developed silicon chips, the number of hardware components has been reduced, which allows the system to maintain high security standards while significantly reducing production costs.
The Evolution of LiDAR and Radar: Challenging the "Long Tail Effect" and Extreme Weather
Satish Jeyachandran, VP of Engineering at Waymo, emphasized that sensors must be resilient to achieve "provable safe AI".
• Lidar:The sixth-generation LiDAR technology has benefited greatly from the cost reduction in the supply chain in recent years. Through a redesigned optical architecture, it has enhanced its ability to handle highly reflective surfaces (such as traffic signs) and can even "see through" the thick water spray from vehicles ahead on highways.
• Imaging Radar:Through the new algorithm, the size and speed of objects can be tracked in real time, providing necessary backup in environments where vision is obstructed, such as icy and snowy conditions.
More than just a car, it's a "Driver": a universal ambition across platforms.
This time, Waymo reiterated a key point: they are developing "Driver," not "Vehicle." This sixth-generation system is highly versatile, allowing for rapid adjustments to sensor configurations and generalization of AI models, whether for Waymo's own "Ojai" platform or the later IONIQ 5 developed in collaboration with Hyundai.
Currently, Waymo's autonomous driving factory in Phoenix is moving towards a scale of "tens of thousands of units per year". By collaborating with OEM partners, it ensures that the base models have the "Waymo Driver Ready" specification when they leave the factory. This is the starting point for the second half of the commercialization of autonomous driving - the "economic scale war".
Analysis: The final battle between pure vision and sensor fusion?
While Tesla was still sticking to a pure vision solution and trying to push the limits through algorithms, Waymo chose a more "resourceful" path: using technological advancements to drive down the price of expensive sensors, while using multi-module backup to handle that "one in a million" unexpected event.
For users, the most noticeable benefit of the sixth-generation system will likely be that Waymo will be as reliable as it is under the California sun, whether in torrential rain, heavy snow, or in dark, remote areas. This cost reduction will ultimately be reflected in the pricing of the ride-hailing service.
