Lidar
Also known as: LiDAR, laser ranging, 3D scanning, point cloud, time-of-flight sensor, depth sensor
Lidar (Light Detection and Ranging) is a remote sensing technology that measures distance by emitting laser light pulses and measuring the time for reflected light to return.
By rapidly scanning in multiple directions, lidar creates detailed 3D point cloud maps of the surrounding environment -- showing the precise position of every surface within range. Lidar is used in autonomous vehicles (detecting obstacles and mapping road geometry), robotics (3D environment mapping for navigation), geospatial surveying (terrain mapping from aircraft), and, in a miniaturized form, in consumer devices (iPhone and iPad Pro depth sensors for AR and camera autofocus).
Lidar, explained at three levels
At home
Lidar works like radar but with light instead of radio waves. The sensor fires extremely fast pulses of laser light in all directions around it, thousands of times per second. When the laser pulses hit nearby objects (cars, pedestrians, trees, buildings), they bounce back. By measuring how long each pulse takes to return (the 'time of flight'), the sensor can calculate exactly how far away each surface is. Do this in all directions simultaneously and you get a detailed 3D map of everything surrounding the sensor -- like a video game world but made of millions of precise distance measurements. This lets self-driving cars 'see' their environment in three dimensions.
At work
Lidar creates 3D point cloud data -- essentially a digital model of the physical world with millimeter-level accuracy. Business applications include: autonomous vehicle navigation (the dominant sensing technology for most AV systems), geospatial mapping (surveying land, infrastructure inspection, forest management), robotics (warehouse robots navigating dynamic environments), construction (as-built documentation comparing designed vs. actual dimensions), and consumer devices (iPhone Pro's lidar enables AR apps and faster camera autofocus). The data format (point clouds) is processed by specialized software (ROS for robotics, SLAM algorithms for mapping, AutoCAD/Revit for construction).
At home
Lidar point clouds: each laser pulse return is recorded as a point in 3D space with coordinates (X, Y, Z) and sometimes intensity (how much light was reflected). Thousands of pulses per second across a rotating scanner create a dense 3D point cloud -- potentially millions of points representing all visible surfaces. Processing this data requires: segmentation (grouping points that belong to the same object), object detection (classifying segments as car, pedestrian, cyclist, etc.), tracking (following objects across successive point cloud frames), and map building (combining many frames into a persistent 3D map). These algorithms are computationally intensive, requiring GPU processing for real-time autonomous vehicle applications.
At work
Spinning vs. solid-state lidar: first-generation vehicle lidar (Velodyne) used mechanically rotating assemblies spinning at 10-20 Hz to create 360-degree coverage. Rotating lidar is accurate and has wide field of view but is expensive ($75,000 for early units), mechanically fragile, and difficult to manufacture for automotive mass production. Solid-state lidar (no moving parts) uses MEMs mirrors, optical phased arrays, or flash lidar to scan without rotation -- lower field of view but dramatically lower cost ($100-500 at scale), higher reliability, and surface-mount manufacturing compatibility. Most next-generation AV production vehicles (Volvo EX90, Mercedes S-Class, Lucid Air) use solid-state lidar from companies like Luminar, Innoviz, and Cepton.
At home
FMCW lidar as the next generation: traditional time-of-flight lidar measures the travel time of discrete laser pulses. FMCW (Frequency-Modulated Continuous Wave) lidar continuously emits laser light with a linearly changing frequency (chirped signal) and mixes the outgoing and returning signals to extract range from the frequency difference and velocity from the Doppler shift. FMCW lidar provides direct velocity measurement (not just position), has better performance in sunlight and adverse weather (continuous signal vs. pulsed), and is more immune to interference from other lidar systems -- important as AV deployments scale. Companies including Aeva, Aurora, and Waymo (in-house) are developing FMCW lidar for next-generation AV sensing.
At work
Camera-only vs. lidar for autonomous vehicles (the Tesla vs. Waymo debate): Tesla's approach to autonomous driving (Autopilot/FSD) uses cameras only (no lidar), relying on neural networks to estimate 3D geometry from 2D camera images and accumulated fleet learning from millions of vehicles. Waymo, Cruise, and most traditional AV developers use lidar as a primary sensor, arguing that cameras cannot reliably estimate absolute distances in all conditions. The debate is partly philosophical (Tesla: cameras are what humans use; Waymo: humans would use lidar too if they had it) and partly empirical (Tesla has more on-road data by orders of magnitude; Waymo has higher demonstrated reliability per mile in operational domains). Camera-only approaches face fundamental physics constraints on depth estimation at distance; lidar provides direct, accurate range measurement regardless of environmental texture.
Real-world examples
At home
Fatima used an iPad Pro with lidar to scan her apartment before a furniture purchase. She used a room measurement app that had the lidar create a 3D model of her living room in about 30 seconds. She then placed virtual furniture in the scan to visualize how it would look and whether dimensions fit. The lidar-based scan was dramatically more accurate than measuring by hand -- it captured the exact room dimensions including alcoves and doorways. She avoided buying a sofa that would have blocked a heating vent because the 3D scan showed it clearly.
At work
Sophie managed surveying for a civil engineering firm. Traditional land surveying required crews of two or three spending days at a site with total stations. She adopted mobile lidar (a backpack unit with scanners and GPS) that one surveyor could operate. The mobile lidar captured a complete 3D point cloud of a 2-km road corridor in 4 hours; the point cloud contained more accurate spatial data than two weeks of traditional survey work. The as-built point cloud integrated directly with the design software, enabling precise comparison of existing vs. planned infrastructure conditions.
Common confusions
- Many people assume lidar and radar work the same way and are interchangeable for autonomous vehicles — Lidar uses laser light (wavelength 905nm or 1550nm) and radar uses radio waves (77GHz for automotive), giving them fundamentally different performance profiles. Lidar produces extremely dense, high-resolution 3D point clouds (centimeter-level precision, millions of points per second) but is degraded by heavy rain, fog, and snow (light scatters in precipitation). Radar has much lower resolution (cannot distinguish fine object details) but penetrates fog, rain, and snow and directly measures velocity (Doppler effect). Most robust AV sensor stacks combine both: lidar for high-resolution environment mapping in good conditions, radar for reliable object detection and velocity in adverse weather. Cameras provide color and texture information that neither lidar nor radar captures.
- Some assume the iPhone's lidar sensor works like automotive lidar and enables full 3D environment mapping for apps — iPhone/iPad Pro lidar uses a flash lidar design (emits a wide diffuse laser pulse rather than a scanning beam) with limited range (approximately 5 meters effective range for face/room scanning vs. 200+ meters for automotive lidar). It is designed for AR overlays (placing virtual objects on real surfaces), camera autofocus (rapid depth estimation for portrait mode and Night mode), and room-scale AR experiences -- not for the long-range, 360-degree, high-resolution point cloud generation needed for autonomous vehicles. The consumer device uses a 9x9 or 16x16 pixel depth array, while automotive lidar generates millions of points per second across a full hemisphere.
For analysts
Lidar awareness is relevant in advisory contexts involving autonomous systems, spatial computing, and geospatial data. Practical areas: autonomous vehicle technology assessments (which sensor suite is appropriate?), construction and infrastructure (point cloud-based as-built documentation is becoming standard), facility management (3D scans for building information modeling updates), agriculture (drone lidar for crop health and yield estimation), and forestry (aerial lidar for biomass estimation and fire risk assessment). The key business consideration is the point cloud data pipeline: lidar sensors generate enormous amounts of raw data that require specialized processing software, storage infrastructure, and trained operators to extract value -- the sensor itself is only the first component.
History & origin
Lidar technology was developed in the 1960s, initially for meteorology (measuring atmospheric particle density) and lunar ranging (measuring the Earth-Moon distance using retroreflectors left by Apollo missions). Survey lidar mounted on aircraft began mapping terrain from the 1980s. Stanford's Stanley (DARPA Grand Challenge winner, 2005) popularized vehicle-mounted lidar by demonstrating autonomous off-road driving. Velodyne (founded 2005) commercialized spinning 360-degree lidar for automotive use; their HDL-64E became the standard sensor on most early autonomous vehicle prototypes. Waymo (Google's self-driving project), Cruise, and dozens of AV startups built fleets using roof-mounted spinning lidar. Competing approaches (solid-state lidar, FMCW lidar) emerged in the 2010s to reduce cost and improve reliability for mass production. Apple integrated miniature lidar (VCSEL-based) into the iPhone 12 Pro (2020) and iPad Pro for AR applications and improved Night mode portrait photography.