Over 30 definitions of key LiDAR terms and concepts from robotics, 3D mapping and autonomous perception.
Angle between two adjacent laser beams, expressed in degrees or milliradians. Fine angular resolution (0.05°) produces a dense point cloud with more detail but reduces coverage at equal range. It depends on channel count and FoV. Example: a 128-channel LiDAR over 18° vertical = 0.14° vertical resolution.
Number of independent lasers in a 3D LiDAR, each emitting a beam at a different vertical angle. More channels produce a vertically denser point cloud. A 16-channel LiDAR offers coarse vertical coverage, a 128-channel provides fine resolution. Examples: Ouster OS0 128 channels, RoboSense Helios 32 channels.
Virtual replica of a physical asset (building, factory, city, infrastructure) updated in real time from sensor data. LiDAR is a key sensor for creating the digital twin by providing accurate 3D geometry of the asset. Used in Industry 4.0, infrastructure management and smart cities.
The angular coverage of a LiDAR sensor, typically expressed in degrees for horizontal and vertical axes. A 360° horizontal FoV is typical of spinning LiDARs, while solid-state sensors offer 90° to 120°. Vertical FoV ranges from 15° to 90° depending on the model and determines height coverage.
LiDAR technology that measures both distance and Doppler velocity by modulating the laser frequency. Unlike ToF, FMCW measures the frequency shift between emitted and received signals, enabling direct velocity detection of objects without post-processing. Less sensitive to sunlight. Examples: Aeva, SiLC, Mobileye (project).
Number of complete point clouds produced per second by the LiDAR. Typically 10–20 Hz for mobile LiDARs, 5–10 Hz for spinning sensors, up to 100 Hz for some proximity sensors. A higher rate enables better fast-object detection but generates more data to process.
Common field of view area between multiple LiDARs mounted on the same vehicle or robot. Good overlap enables better detection in blind spots, improves redundancy and increases point density in the critical zone. Typically used in autonomous vehicles (4-6 LiDARs) or mobile robots (2 opposing LiDARs).
Combination of a GNSS receiver (GPS, GLONASS, Galileo) and an inertial navigation system (IMU) used to geolocate mobile mapping LiDAR. GNSS provides absolute position, INS bridges signal loss (tunnels, dense urban areas). LiDAR data is then aligned with the GNSS/INS trajectory to produce georeferenced point clouds.
LiDAR generates point clouds dense enough to reconstruct human silhouettes, license plates or private interior details. In Europe, this may qualify as personal data under GDPR. Intrusion detection systems using LiDAR must implement source anonymization (aggregation, downsampling, zone blurring).
IEC 60529 standard rating a sensor's resistance to dust (first digit, 0-6) and water (second digit, 0-9). For outdoor LiDAR, IP67 (dust-tight + 1m/30min immersion) or IP69K (high-pressure jets) is recommended. Mobile LiDARs range from IP54 (splash protected) to IP69K (intensive washdown).
A sensor composed of accelerometers and gyroscopes measuring linear acceleration and angular velocity. Used with LiDAR to compensate for carrier motion, improve SLAM accuracy, and provide pose estimation between LiDAR acquisitions. Essential in mobile mapping and mobile robotics.
Safety classification of the LiDAR laser per IEC 60825 standard. Classes range from 1 (eye-safe without precautions) to 4 (immediate danger). Most consumer LiDARs are Class 1. Long-range LiDARs may be Class 1M (hazardous if viewed with optical instruments) or Class 3R (low risk).
Intrinsic calibration corrects the sensor's internal defects (measurement bias, beam alignment). Extrinsic calibration determines the LiDAR's position and orientation relative to the vehicle or other sensors (camera, IMU). Accurate calibration is essential for sensor fusion and georeferencing.
Spurious points in a LiDAR point cloud, caused by multiple reflections, ambient sunlight, rain, snow, fog or inter-sensor interference. Noise appears as floating points, artifacts or reduced accuracy. Temporal, statistical and geometric filters can reduce it.
Semi-solid-state LiDAR using a microscopic oscillating mirror (MEMS) to steer the laser beam. A compromise between robustness and performance: fewer moving parts than spinning mechanical, but FoV and resolution limited by mirror size. Examples: InnovizOne, RoboSense M1, Cepton.
A LiDAR's ability to detect multiple echoes from a single laser pulse. Useful in vegetated environments: the first echo may come from a tree canopy, the second from an intermediate branch, the third from the ground. Single-return LiDARs only capture the first echo and lose vegetation cover information.
A set of points in 3D space generated by a LiDAR, each point having coordinates (x, y, z) and sometimes intensity or color. The point cloud is the raw data produced by the sensor, before any processing such as classification, segmentation or 3D reconstruction.
IEEE 1588 time synchronization protocol used to precisely align data from multiple LiDAR sensors with each other or with other sensors (IMU, camera, GNSS). Accuracy reaches microseconds, essential for sensor fusion in robotics and autonomous vehicles.
Number of LiDAR points per square meter on a scanned surface. Depends on range, angular resolution, refresh rate and carrier speed. For mobile mapping, aim for 100-500 pts/m². For infrastructure inspection, 1000+ pts/m². Insufficient density makes detection of fine objects impossible.
The process of automatically labeling points in a cloud into categories (ground, building, vegetation, vehicle, pedestrian, etc.). Uses machine learning or deep learning algorithms (PointNet, RandLA-Net, KPConv). Essential for autonomous navigation, mapping and scene analysis.
Division of a point cloud into distinct regions or objects (instance segmentation) or semantic categories (semantic segmentation). Not to be confused with classification: segmentation identifies groups of points belonging to the same object (e.g., this vehicle, this pedestrian).
Main file formats for storing LiDAR point clouds. PCD (Point Cloud Library): ROS format, simple, open. LAS/LAZ: geospatial industry standard, with classification, return, intensity. PLY: polygonal format, used for 3D meshes and export. E57: exchange format for 3D data with images. Format choice depends on application (ROS → PCD, geomatics → LAZ, visualization → PLY).
Precision measures repeatability: if the same distance is measured 100 times, are the values close to each other? Accuracy measures error relative to the true value. A LiDAR can be precise (low noise) without being accurate (5 cm systematic offset). Datasheets usually state accuracy (e.g., ±2 cm at 50 m).
A surface's ability to reflect the laser signal. Expressed as a percentage (10%, 80%…), reflectivity directly affects the effective LiDAR range. A stated range of '300 m @ 10% reflectivity' means the sensor detects an object with 10% reflectivity up to 300 m — range will be much greater on a highly reflective white surface.
Open-source framework for robotics software development. ROS2 (Humble, Iron, Jazzy) is the modern version, offering pub/sub communication, drivers for many sensors (including LiDAR), visualization tools (RViz2) and SLAM, perception and navigation libraries. ROS support is a key criterion for LiDAR integration in robotics.
A technique for simultaneous localization and mapping that allows a robot or vehicle to build a map of its environment while localizing itself in real time. SLAM uses LiDAR (or camera) data combined with odometry and sometimes GNSS to estimate trajectory and map simultaneously.
LiDAR technology where the sensor physically rotates 360° to scan the environment. This is the most mature technology (Velodyne HDL-64, Ouster OS, Hesai Pandar). It offers 360° horizontal FoV but has moving parts that limit lifespan and vibration robustness.
LiDAR with no moving parts, using an array of photodetectors (flash) or an optical phased array (OPA) to steer the beam. Robust, compact and potentially low-cost at scale, but limited FoV (usually 90-120°). Examples: Hesai FTX, Blickfeld Cube, LeddarTech.
Combination of data from multiple sensors (LiDAR, camera, radar, IMU, GNSS) to obtain more robust and complete perception than any single sensor. LiDAR provides precise 3D depth, camera provides color and semantic classification, radar provides velocity and range in bad weather.
Distance measurement principle using time of flight: the LiDAR emits a laser pulse and measures the time it takes to return after reflecting off an object. This is the most common method for mechanical, MEMS and flash solid-state LiDARs. Distance is calculated as d = c × t / 2 where c is the speed of light.
The laser wavelength used by the LiDAR, typically 905 nm (near-infrared) or 1550 nm (infrared). 905 nm is cheaper and more compact but less powerful and potentially eye-hazardous at high power. 1550 nm is more eye-safe, performs better in fog/rain, but requires more expensive components (InGaAs lasers).