Solid-state lidar: laser distance sensing without spinning parts

Lidar works by firing laser light at a scene and timing how long it takes to bounce back. Repeat that across many directions and the sensor builds a 3D map of its surroundings. The standard design spins to give a 360-degree view. Solid-state lidar has no moving parts, or in some designs only tiny ones (Wikipedia's entry on lidar). Supporters expect this to make sensors smaller, tougher and cheaper to build in volume. That matters for cars that need to see in the dark and for robots that must avoid people and objects. Most of the evidence so far comes from laboratory chips and company announcements. This article separates what has been shown from what is only claimed.

How the main designs work

Solid-state designs fall into a few families.

• Flash lidar lights the whole field of view with one wide pulse and reads the return on a detector array, so it works like a camera that takes pictures of distance (Wikipedia's entry on lidar).
• Optical phased arrays split laser light between many emitters on one chip and adjust the phase of each. Changing the phase differences between the emitters forms and steers the beam, with no moving parts (Wikipedia's entry on phased-array optics).
• FMCW lidar uses frequency-modulated laser light. It can measure a target's radial velocity directly as well as its distance, and it can be built on a chip, as in the Pointcloud work below (Pointcloud dataset, Dryad).

Micro-mirror (MEMS) systems still move, but at a tiny scale, so they sit between mechanical and fully solid-state designs (Wikipedia's entry on lidar).

What has been demonstrated

Laboratory phased arrays. A paper in Optics Letters (online 20 December 2016, issue dated January 2017) reported passive large-scale phased arrays made with silicon nitride on a CMOS-compatible silicon photonic platform. At 1550 nm, an array with a 4 mm by 4 mm aperture produced what the authors call "a record small and near diffraction-limited spot size" of 0.021 degrees by 0.021 degrees. The abstract also reports 10 dB of side lobe suppression and a main beam power of 400 mW. Because the arrays are passive, the abstract does not report steering the beam. It does not describe ranging tests, and it presents the devices as chip demonstrations rather than working sensors for vehicles (Poulton and colleagues, Optics Letters, 2017).

What is only reported

Company and press figures show scale, but they are not independent tests.

• Chip sensor dataset (company-authored). A dataset published on Dryad on 13 February 2026, whose listed authors are all at Pointcloud GmbH, describes a coherent FMCW focal plane array with 352 by 176 pixels and over 0.6 million photonic components on the chip. It shows point clouds of objects 4 to 65 m away, with per-pixel integration times compatible with frame rates of 3 to 15 fps, and measures radial velocity directly. The record does not state target reflectivity, weather or field of view for those distances, and it does not link a peer-reviewed paper. It is a useful signal but not an independent test (Pointcloud dataset, Dryad).
• Car volumes. In its unaudited full-year 2025 results, dated 24 March 2026, Hesai reported 1,381,133 lidar units shipped for driver-assistance (ADAS) applications in 2025, up from 456,386 in 2024. The release links its ADAS business to passenger and commercial vehicles. It does not say how many units are solid-state, or how many went into private cars (Hesai full-year 2025 results, SEC filing).
• Robot volumes. In results announced on 25 March 2026, RoboSense reported about 912,000 lidar units sold in 2025, up 67.6%, of which about 303,000 went to robotics. The release does not give a driver-assistance unit figure, and it does not say which lidar designs were sold (RoboSense 2025 results, PR Newswire).
• Price. A Nikkei Asia article republished by KrASIA on 15 November 2025 reports that Hesai has lowered the price of its new ATX lidar to USD 200. The same article says lidar typically costs over USD 10,000 per unit, and quotes Li Yifan, Hesai's co-founding chief executive: "Unless we significantly reduce the price, it will be difficult to install it in all vehicles, and safety will not improve." The USD 200 figure comes from the article; Hesai's results release mentions the ATX but gives no price (KrASIA).

Car uses

The volume figures above are mostly for driver assistance, not self-driving. Supplier risk is clear in the case of Luminar, a lidar company that supplied Volvo. Bloomberg reported that Luminar filed for Chapter 11 bankruptcy protection on 15 December 2025 in the Southern District of Texas. Luminar said in a filing on 17 November 2025 that it had received notice that Volvo had ended a supply chain relationship for its lidar products (Transport Topics, republishing Bloomberg).

Robot uses

Robots take a sizeable share of some lidar volumes, as the RoboSense figures show, but the sources cited here do not measure how solid-state designs perform on robots. Speculative: the case for them rests on size and robustness, and that has not been tested in the sources cited here. Wikipedia notes that MEMS mirror systems "can be disrupted by shock/vibration and may require repeated calibration", which matters for vehicles and robots alike (Wikipedia's entry on lidar).

Limits and open problems

• Range and weather. The only range figure here is the 4 to 65 m in the company-authored dataset, which is not an independent test. The laboratory phased array result gives no range. Wikipedia notes that heavy rain can add noise to lidar data, but we found no verified figures for long range, or for performance in rain, fog, snow or dust.
• Eye safety. Laser power is limited for eye safety. Wikipedia says some systems cap their laser power or use an automatic shut-off at certain altitudes, and that 1,550 nm lasers are eye-safe at relatively high power because water strongly absorbs that wavelength. Seeing at a distance needs a powerful burst of light, so the trade-off matters for range. We have not verified which safety standard applies, so this article does not name one.
• Cost. There is no verified unit cost for a solid-state car sensor. The USD 200 price is a reported figure.
• Supplier risk. The Luminar case shows that one supplier's failure can disrupt a carmaker's plans.
• Interference and standards. We found no verified source on how lidar units interfere with one another, or on a shared test standard for comparing sensors.

What to watch next

• Whether Hesai and RoboSense report passenger car volumes separately from commercial and robot units.
• Peer-reviewed publication of the Pointcloud work, with test conditions for range, reflectivity and weather.
• Independent range tests in rain, fog and against dark targets.
• Unit prices for car lidar published in company filings rather than press reports.
• Any agreed test standard, or any change to eye-safety power limits.