Hesai explains how lidar 3D perception works, from point clouds to full-color sensing

Lidar, short for light detection and ranging, emits laser pulses and measures how long the reflected light takes to return; because the speed of light is known, it can calculate the distance to the surface that reflected each pulse, and repeating this many times per second builds a 3D representation called a point cloud. The technology gives machines a direct, three-dimensional understanding of the space around them and is increasingly used in robotics, autonomous systems, mapping, industrial automation and infrastructure. The sensor supplies measurements rather than decisions, so a robot may use the data to identify free space and avoid obstacles while a mapping system reconstructs a building or industrial site. Depending on the sensor and data format, individual points can also carry intensity, reflectivity and timestamps, which software can use to classify objects, reconstruct movement, measure changes or support navigation. Object recognition and classification, such as reading a group of points as a person, vehicle, pallet or wall, are normally performed by that connected software.

Lidar emits its own light, so it does not depend on daylight or artificial lighting and can perceive its surroundings in complete darkness, while modern lidar is also designed for bright environments using optical filtering, sensitive receivers and signal processing to separate reflected laser signals from strong ambient light. Hesai's Intelligent Point Cloud Engine, used in products such as the OT128 and ATX, identifies rain, fog, dust, exhaust fumes and water spray and filters environmental noise in real time, which Hesai says supports clearer all-weather perception for outdoor robots, autonomous vehicles, infrastructure and industrial systems. Lidar also detects dark objects such as tires, dark clothing or asphalt because it measures reflected laser light rather than visible color, which is why range is often stated together with a reflectivity value, such as range at 10 percent reflectivity. A rotating lidar uses moving scanning elements for broad horizontal coverage, often a complete 360-degree view, while a solid-state lidar covers a defined field of view without a continuously rotating external assembly, and hybrid architectures combine optical, electronic and mechanical elements. No approach is automatically better: the right architecture depends on the required field of view, range, resolution, form factor and installation position.

Hesai says lidar is inherently privacy-friendly because it measures spatial information rather than capturing conventional images, so a point cloud does not capture facial features, identities or other personally identifiable visual details, though customers should assess applicable privacy requirements for colorized point clouds. Hesai lidar sensors do not store point-cloud data, not even a second of it, and cannot transmit it wirelessly because they have no cellular connection, Wi-Fi or Bluetooth; data is transferred through a secure wired connection to the customer's computing system, and the customer retains ownership and control. Hesai publishes Class 1 eye-safety status for models including the OT128, ATX and JT series, meaning the laser is considered eye-safe under the product's defined operating conditions, and the company notes the sensor should be installed and operated according to the relevant product documentation. For developers, Hesai provides product documentation, downloadable resources and point-cloud tools, including PandarView and PandarView 2, which allow users to visualize live point clouds, record and replay data, inspect individual points and export selected data. Hesai adds that lidar can support localization and mapping without GPS by comparing live point-cloud data with a map or creating one as the system moves, that it cannot see through solid walls, and that no single specification determines the best sensor: range, field of view, resolution, accuracy, point density, size, power consumption and environmental performance should be considered together based on the application.