Automotive

LiDAR

LiDAR — Light Detection and Ranging — measures distance the same conceptual way as radar, timing a round trip, but with laser light instead of radio waves. The much shorter wavelength buys far higher precision and spatial resolution, at the cost of light being far more easily blocked by fog, heavy rain, or a dirty lens than radio ever is.

Quick reference

What it is: An optical sensor that measures distance by timing how long a laser pulse takes to reflect back, building a precise 3D point cloud of the surroundings.
Mental model: Radar's cousin, but using light — much finer "ruler" for distance and angle, much more easily blocked by anything the light can't pass through.
Key numbers: Automotive LiDAR typically ranges 100–250 m, with angular resolution down to a fraction of a degree and hundreds of thousands of points measured per second.
Reach for it when: You need precise 3D shape and position — not just "something is there" — for mapping or a high-resolution perception stack.

Time-of-flight: the core measurement

A LiDAR emits a very short laser pulse and starts a timer. The instant that pulse reflects off something and comes back, the timer stops. Distance is just the classic distance = speed × time relationship, halved because the light makes a round trip:

d = (c × t) / 2

For an object 30 m away, t ≈ 200 ns — LiDAR timing electronics need sub-nanosecond precision to turn that into useful centimeter-scale range resolution.

Some LiDAR designs measure phase shift of a continuously modulated beam instead of timing discrete pulses (conceptually closer to how FMCW radar works), trading some peak range for better resolution at short range — but pulsed time-of-flight is by far the dominant approach in automotive LiDAR today.

Scanning and the point cloud

One pulse gives one distance measurement in one direction. To build up a useful picture of the surroundings, a LiDAR rapidly repeats this across many different angles — steering the beam mechanically, with a tiny mirror, or by other means — producing a dense set of 3D points, each with its own measured distance:

A production automotive LiDAR measures hundreds of thousands of these points per second, across a wide field of view — dense enough to reconstruct the 3D shape of nearby objects, not just their existence.

Mechanical, MEMS, flash, and OPA LiDAR

Type	How it scans	Notes
Mechanical (spinning)	The entire laser/detector assembly physically rotates	Mature, wide (often 360°) field of view; moving parts are a cost and long-term reliability concern in a vehicle.
MEMS	A tiny mirror steers the beam electronically	No large rotating assembly — smaller, more automotive-friendly packaging than mechanical units.
Flash	Illuminates the whole scene at once, like a camera flash	No moving parts and no scanning at all; trades range for solid-state simplicity and a fully "instant" full-frame capture.
OPA (optical phased array)	Steers the beam electronically by controlling light phase across an array	Fully solid-state beam steering — the direction furthest from a spinning mechanical unit, still maturing for automotive volume production.

The LiDAR processing chain

Structurally almost identical to the radar processing chain — emit, bounce, receive, measure, hand off to the ECU — just with light instead of radio.

LiDAR vs. radar vs. camera

See the full comparison table on the Radar page rather than repeating it here — the short version is that LiDAR's strength is precise 3D shape and position, at a cost premium and a real weakness in heavy rain, fog, or snow that radio-based radar doesn't share.

Real-world applications

High-definition mapping — survey vehicles use LiDAR to build the centimeter-accurate 3D maps some autonomous driving stacks localize against.
Autonomous vehicle perception — many robotaxi and higher-autonomy programs rely on LiDAR as a primary sensor precisely for its precise 3D shape data, fused with radar and camera.
Curb and terrain detection — LiDAR's precise height data helps distinguish a curb, a speed bump, or an actual obstacle far more reliably than camera alone.
3D scene reconstruction — the dense point cloud is directly useful for building a geometric model of the environment, not just detecting that objects exist.

Gotchas

Field notes

Heavy rain, fog, snow, or a dirty lens genuinely degrade LiDAR performance — near-infrared light scatters off airborne water and dirt far more than radio waves do, unlike radar.
Very dark, light-absorbing surfaces (matte black paint, for instance) return a much weaker signal — a real, counterintuitive limitation, not just a hypothetical edge case.
Eye-safety regulations cap how much laser power a LiDAR can emit, which directly caps achievable range and return-signal strength — this isn't just an engineering choice, it's a hard regulatory constraint.
Cost remains the biggest adoption barrier relative to radar and camera — a major reason many production ADAS systems still rely on radar+camera fusion alone without LiDAR at all.