Multi-Sensor Architecture as Waymo’s Core Advantage
Waymo Co-CEO Dmitri Dolgov delivered a keynote speech at a Y Combinator event on August 10, offering critical insights into the sensor configuration of autonomous driving technology. Dolgov explained that Waymo’s self-driving vehicle system integrates three types of sensors—cameras, lidar, and radar—each serving distinct advantages: cameras provide high-resolution imagery and color recognition, lidar precisely measures the 3D structure of the environment, and radar maintains stable detection in adverse weather conditions. He stressed that this multi-modal combination effectively overcomes the limitations of single sensors, such as low-light or glare interference.
Dolgov specifically noted that lidar and radar are active sensors capable of maintaining high-precision detection even in pitch-black darkness or blinding sunlight. In contrast, vision-only systems may face significant limitations in nighttime or bright-light conditions. He stated that this multi-sensor architecture not only enhances the system’s environmental adaptability but also reduces safety risks caused by the failure of a single sensor.
Vision-Only Systems’ Safety Performance 'Peaks Too Early'
During his speech, Dolgov explicitly stated that the safety performance improvement curve for vision-only autonomous driving systems 'peaks too early,' making it difficult to achieve the 'superhuman' level required for full self-driving. He pointed out that humans drive using only vision, proving that vision-only systems are feasible for products aiming at 'near-human performance' or 'driver-assistance.' However, if the goal is to achieve full autonomy and surpass human performance, the safety performance of vision-only solutions will struggle to continue improving.
He further explained that while cameras provide rich environmental information, their detection capabilities significantly decline in low-light or glare conditions. Lidar, on the other hand, can precisely measure object distances and shapes in darkness, while radar can penetrate fog, rain, and snow, directly measuring object speed through the Doppler effect. Dolgov argued that the complementarity of these three sensors effectively compensates for their individual shortcomings, enhancing the overall reliability of the system.
Divergence in Autonomous Driving Technology Approaches Emerges
Dolgov’s remarks highlighted the current divergence in autonomous driving technology approaches. While Waymo adheres to a multi-sensor architecture, companies like Tesla continue to push vision-only solutions. Tesla’s Full Self-Driving (FSD) system still relies primarily on cameras and does not incorporate lidar or radar. Despite Tesla’s claims that its vision-only system has made progress in urban road testing, its safety and reliability remain subjects of external skepticism.
Dolgov did not directly name competitors in his speech, but his criticism of vision-only systems 'peaking too early' clearly addressed the ongoing debate over autonomous driving technology paths. He stated that Waymo’s multi-sensor architecture ensures stable performance across various environments, with the advantages of active sensors becoming particularly evident in nighttime or adverse weather conditions. This choice of technology path will directly impact the commercialization process and safety performance of self-driving vehicles.
Waymo currently offers autonomous ride-hailing services in multiple U.S. cities and continues to expand its operational scale. Dolgov’s speech not only demonstrated Waymo’s confidence in its technology path but also provided a critical direction for the future development of autonomous driving. As the commercialization of self-driving vehicles accelerates, debates over sensor configurations are expected to intensify, with each company’s technology choices determining its competitiveness and safety performance in the market.
