Hesai explains how lidar helps warehouse robots, farm machines, drones and humanoids read physical space

A warehouse robot may have a perfectly planned route, until someone leaves a pallet in its path: the destination has not changed, but the space around the machine has. Hesai's argument is that robots therefore need to understand their surroundings as they are now, not just follow instructions based on how those surroundings looked earlier. Lidar measures distances to surrounding surfaces and creates a geometric view of the world, and in 3D systems this forms a point cloud that software uses to estimate position, detect obstacles and judge the space available for movement. Working alongside cameras and motion sensors, it helps a robot turn an understanding of its environment into action.

Each job calls on spatial perception differently. A transport robot must identify a newly blocked aisle and find usable space around it, a forklift must approach racks and pallets with enough clearance for its load, and an inventory robot can use 3D measurements both to navigate and to record the layout of stock and shelving. A mobile manipulator gets a geometric reference as it approaches a work area, while cameras and closer-range sensors support the handling task itself. The value is not simply reaching a destination, Hesai says, but adapting movement and positioning to a physical workspace even when that workspace changes.

Lidar-based localization also reduces dependence on external positioning signals. GNSS signals, including GPS, can be unavailable inside buildings and unreliable beneath dense tree cover, so using lidar measurements as a reference, often together with onboard motion sensors, lets robots navigate without continuous satellite reception — particularly valuable for machines operating across indoor and outdoor environments. Lidar can further contribute to simultaneous localization and mapping, or SLAM, allowing a robot to estimate where it is while building or updating a map. In an orchard, for example, a machine must steer between tree trunks, account for uneven ground and hold its position under foliage that interferes with satellite reception, while cameras help a weeding or spraying robot distinguish crops from weeds.

Drones, quadrupeds and humanoids add a harder problem: they must interpret the space above, below and around their bodies, not just the next stretch of a route. An inspection drone entering a storage tank or industrial structure may have to fly around pipes and beams without satellite navigation, while a humanoid reaching for a shelf or opening a cabinet must judge how much room an arm has to move. Lidar supplies that 3D geometry, cameras help identify objects, and joint sensors and touch feedback support coordinated movement. Hesai ties this perception-to-action link to Physical AI, and notes that navigation and perception lidar is not automatically a certified protective device — safety-related functions require an appropriately designed and validated system.