Most automated-driving developers, including Waymo and most traditional automakers, use sensor fusion — cameras plus radar plus lidar — because each sensor covers the others' weak points. Tesla is the highest-profile holdout, betting on a camera-only approach on the theory that vision plus AI should eventually match human-level perception without extra hardware. Lidar unit costs have fallen sharply over the past decade, from tens of thousands of dollars per unit to figures more in the hundreds to low thousands for automotive-grade units, which has made fusion more affordable and shifted the debate from "can we afford lidar" to "do we need it."
At a glance
| Sensor | Strength | Weakness |
|---|---|---|
| Camera | Reads color, text, signs, lane markings; cheap | Struggles in glare, fog, heavy rain, darkness |
| Radar | Sees through rain/fog; measures speed directly | Low resolution; struggles to classify object shape |
| Lidar | Precise 3D distance mapping, works in the dark | Costlier; can be degraded by heavy fog/snow |
Two philosophies, one goal
Every automated driving system needs to answer the same question many times a second: what's around the car, how far away, and how fast is it moving? Camera-only advocates argue that since humans drive using eyes alone, a sufficiently advanced vision-and-AI system should be able to do the same, and that adding sensors adds cost and complexity without a corresponding safety gain once the software is mature enough. Sensor-fusion advocates argue that redundancy is the point — if a camera is blinded by sun glare or a radar return is ambiguous, lidar's precise depth data can resolve the conflict independently, and multiple sensor types failing in the same way at the same time is far less likely than one type failing alone.
Where each side sits today
- Tesla: camera-only ('Tesla Vision') across its consumer ADAS stack, arguing cameras plus neural networks will match or exceed lidar-based perception as the software matures
- Waymo, most robotaxi operators, and most traditional automakers building Level 2-3 systems: camera, radar, and lidar together, treating each sensor as a check on the others
- Mobileye: primarily camera-led but increasingly offers both camera-only and sensor-fusion product tiers depending on the automation level targeted
Why lidar got cheaper — and why that matters
Early automotive lidar units, the spinning rooftop sensors seen on first-generation robotaxi prototypes, cost tens of thousands of dollars each, which made them impractical for consumer vehicles. Solid-state lidar, with no moving parts, and manufacturing at higher volumes have pushed unit costs down dramatically over the past several years, to the point where several automakers now offer lidar-assisted highway driving features on production vehicles rather than only research prototypes. That price drop is the main reason the debate has shifted — fusion is no longer automatically the expensive option.
What this means if you're comparing vehicles
A camera-only system isn't inherently unsafe, and a lidar-equipped system isn't automatically better — what matters is how the whole stack performs in real-world testing and independent ratings, not which sensors are listed on a spec sheet. Marketing materials rarely explain the trade-off honestly, so treat sensor lists as one input, not the whole picture.
No production ADAS system, camera-only or fusion, is currently rated for full self-driving without supervision. Check the specific SAE level of any feature rather than assuming more sensors means more autonomy.
The likely direction of travel
Expect the middle ground to keep narrowing rather than one approach fully winning. As camera-AI models keep improving and lidar keeps getting cheaper, more automakers are likely to add at least some lidar or high-resolution radar as a low-cost safety backstop even in systems that lean primarily on vision, particularly for features that operate at higher speeds or with less driver supervision.
Frequently asked questions
Is lidar always safer than camera-only systems?
Why does Tesla avoid lidar?
Has lidar gotten cheaper?
What does radar add that cameras can't do alone?
Do robotaxi companies use lidar?
Sources & further reading
Figures, prices and policy details were current at the last-updated date above. Automotive pricing, incentives and regulations change frequently — verify time-sensitive details with the linked primary sources. Read our editorial policy and fact-checking standards.
