LiDAR (light detection and ranging) is an active range sensor that emits laser light, detects the reflection and converts the travel time or frequency shift of the light into distance. By steering the beam or using many beams, it measures thousands to millions of points per second and builds a point cloud that describes the geometry around the sensor.
Among robotics sensors, LiDAR is valued for direct, metric depth measurement that does not depend on ambient light or surface texture. Mobile robots use it for mapping, localization and obstacle detection, autonomous vehicles use it for 3D perception, and safety-rated versions guard the workspace of industrial machines.
How LiDAR measures distance
Most LiDAR sensors measure distance with pulsed time of flight: the distance equals the speed of light multiplied by the round-trip time, divided by two. Because light travels about 30 cm per nanosecond, each nanosecond of round-trip time corresponds to about 15 cm of range, and a target at 100 m returns its echo after roughly 667 ns. Precise timing electronics are therefore the heart of the sensor.
Two other principles are in use. Amplitude-modulated continuous wave sensors compare the phase of a modulated beam with its echo, which works well at short range. Frequency-modulated continuous wave (FMCW) sensors sweep the laser frequency and mix the echo with a reference, which also yields the radial velocity of each point through the Doppler effect.
| Principle | What is measured | Typical range | Main strength |
|---|---|---|---|
| Pulsed time of flight | Round-trip time of a short pulse | From a few meters to over 200 m | Simple, mature, long range |
| AMCW phase shift | Phase delay of a modulated beam | Mostly under about 50 m | High precision at short range |
| FMCW | Beat frequency between echo and reference | Long range, comparable to pulsed | Per-point velocity, low interference |
The laser wavelength shapes the design. Sensors at 905 nm use inexpensive silicon detectors but must limit pulse energy to stay eye-safe. Sensors at 1550 nm may emit more power at the same Class 1 eye-safety rating under IEC 60825-1, because the eye absorbs this wavelength before it reaches the retina, yet they need costlier InGaAs detectors and lose more signal to water in rain and fog.
Scanning architectures of LiDAR sensors
A LiDAR sensor’s scanning architecture determines its field of view, point density, durability and price. The beam can be steered mechanically, by a micro-mirror, by solid-state optics, or not at all, in which case a flash of light illuminates the whole scene at once.
| Architecture | How it scans | Typical field of view | Typical robotic use |
|---|---|---|---|
| 2D spinning scanner | One beam on a rotating mirror | Single plane, 180-360 degrees | Indoor AMRs, safety laser scanners |
| Mechanical multi-beam | 16-128 lasers rotating together | 360 by about 30-90 degrees | Outdoor robots, mapping, research vehicles |
| MEMS mirror | Tiny oscillating mirror steers the beam | About 60-120 by 20-30 degrees | Automotive and compact robots |
| Flash LiDAR | Whole scene lit, detector array images it | Narrow to medium, short range | Docking, manipulation, close range |
Spinning units provide the full 360-degree view that many navigation stacks expect, but their moving parts add wear. Solid-state designs trade field of view for robustness, so vehicles often combine several of them to cover all directions.
Key LiDAR specifications
The datasheet values of a LiDAR sensor are meaningful only together with their conditions, and the most important condition is target reflectivity. A sensor quoted at 200 m on a highly reflective target may reach far less on dark asphalt or a black car, so serious specifications state the range at about 10% reflectivity.
| Specification | Typical values | Why it matters |
|---|---|---|
| Maximum range | About 10-30 m indoor, 100-300 m automotive | Detection distance and reaction time |
| Range accuracy | About ±1-3 cm | Map quality and localization precision |
| Angular resolution | About 0.1-0.4 degrees | Smallest object detectable at distance |
| Frame rate | About 10-20 Hz for 3D, up to about 50 Hz for 2D | Latency and motion distortion |
| Point rate | Tens of thousands to several million points per second | Scene density and computing load |
Angular resolution limits what the sensor can see far away. With 0.2 degrees between beams, neighboring points at 20 m lie about 7 cm apart, so a thin pole or a cable may fall between them. For this reason, robots that must detect small obstacles at speed often add a second sensor with a different viewpoint.
LiDAR compared with other range sensors
LiDAR competes with the depth camera, radar and ultrasonic sensors, and each wins in a different regime. Robust robots usually combine two or more of them and merge the data through sensor fusion.
| Sensor | Range and precision | Strength | Main weakness |
|---|---|---|---|
| LiDAR | Long range, centimeter precision | Accurate geometry, independent of lighting | Cost, rain, fog and dust |
| Depth camera | Short range, error grows with distance | Dense depth with color, low cost | Sunlight, featureless surfaces |
| Radar | Long range, coarse angular resolution | Works in rain, fog and darkness | Sparse points, little shape detail |
| Ultrasonic sensor | Under a few meters, coarse | Very cheap, detects glass | Wide beam, slow updates |
Limitations and failure modes of LiDAR
LiDAR fails in predictable ways that follow from its optics. Glass, mirrors and polished metal reflect the beam away or create phantom points behind the surface; very dark materials return too little light; and rain, fog, dust and exhaust scatter the beam and produce false close-range echoes. Many sensors report multiple returns per pulse so that software can separate a raindrop from the wall behind it.
Two further effects matter on moving robots. A spinning sensor captures each frame over 50-100 ms, so a turning robot sees a skewed scan unless the points are corrected with odometry or an IMU, a step called motion compensation or deskewing. When several robots share a space, crosstalk between LiDARs can inject spurious points, which coded pulses and FMCW designs largely suppress.
The idea in brief
LiDAR works like a very fast tape measure made of light. It sends out short flashes of laser light and measures how long each one takes to bounce back. Since light moves at a known speed, the time tells the distance.
By repeating this hundreds of thousands of times per second in many directions, the sensor draws a 3D outline of everything around the robot. The robot uses that outline to know where it is and what is in its way, even in complete darkness.
A practical scenario
An autonomous mobile robot carries pallets through a warehouse at up to 1.5 m/s. Near the floor, a safety-rated 2D laser scanner, certified as electro-sensitive protective equipment under IEC 61496-3, watches a protective field in front of the robot. With a system response time of 0.1 s, the robot travels 0.15 m before braking even begins; adding about 0.4 m of braking distance and a 0.1 m safety supplement gives a protective field of about 0.65 m.
The same scans serve navigation. The robot matches each scan against a stored map of racks and walls and localizes itself to within a few centimeters, while a second, 3D LiDAR on the mast detects overhanging forks and shelf edges the floor scanner cannot see. The outcome is a vehicle that stops reliably for a worker stepping into its path and still docks precisely at each pallet station.