Waymo co‑CEO Dmitri Dolgov didn’t mince words when he argued that camera‑only self‑driving systems simply cannot deliver true autonomy. After riding in multiple robotaxis myself — including Waymo’s and camera‑heavy competitors — I have to agree. The gap isn’t subtle. It’s structural.
Dolgov laid out the reasoning clearly: Lidar and radar aren’t optional add‑ons; they’re foundational to how a fully autonomous vehicle perceives the world. Cameras excel at recognizing objects, reading signs, and interpreting visual cues. But they struggle with depth, distance, and adverse conditions. Rain, fog, glare, low light, and occlusions all degrade camera performance in ways software alone can’t fully compensate for.
That’s where Lidar and radar step in.
- Lidar provides precise 3D mapping and distance measurement, independent of lighting.
- Radar penetrates fog, dust, and darkness, tracking motion with reliability cameras can’t match.
Together, they create a sensor fusion stack that can detect, classify, and predict behavior with redundancy — the kind required for vehicles that must make life‑or‑death decisions without human intervention.
Dolgov emphasized that Waymo’s real‑world testing backs this up. Billions of autonomous miles have shown that edge cases — the unpredictable, rare, chaotic moments — are where camera‑only systems fail most dramatically. A child darting behind a parked car, a cyclist emerging from glare, a truck blending into a bright sky: these are scenarios where Lidar and radar provide the missing certainty.
After experiencing different robotaxis firsthand, the contrast is obvious. Camera‑only vehicles feel reactive, sometimes hesitant, occasionally confused. Waymo’s multi‑sensor approach feels anticipatory — like it’s seeing more than the human eye can.
The debate isn’t philosophical. It’s practical. If the goal is true Level 4 or Level 5 autonomy, Dolgov’s argument stands: cameras alone won’t get us there. A layered sensor suite isn’t just better — it’s necessary.