On July 9, RoboSense officially announced that its total LiDAR sales in the first half of 2026 reached 719,200 units, a year-on-year increase of 169.6%. Among them, LiDAR sales in the robotics sector reached 282,600 units, surging 510.4% year-on-year. In the second quarter of 2026, total LiDAR sales reached 388,900 units, with robotics sector LiDAR sales at 97,100 units, up 182.3% year-on-year.
These figures clearly demonstrate that the market growth of LiDAR has shifted from the past single "automotive track" to a new dual-track paradigm running in parallel with automotive and robotics, making robotics the largest incremental market for the LiDAR industry.
Why 3D LiDAR Is Taking Off
I. The Industry's "Painful Realization"
If you were an early user of robot vacuums, you must have seen their rather clumsy behavior: cleaning diligently while constantly bumping into walls or getting stuck on steps, unable to free themselves. This gives you a good idea of why the robotics industry is so obsessed with giving robots "eyes."
Before the advent of 3D LiDAR, the "vision" of robots went through several stages. The most primitive was the "blind guessing" stage, relying on buried electromagnetic wires to move in circles; once out of the boundary, they lost direction. Later, GPS and cameras were introduced, but they immediately "lost themselves" upon entering spaces like hallways. Cameras were also constrained by lighting, becoming useless in the dark. Then 2D LiDAR became popular. It could tell the robot "there is something ahead," but could not distinguish what it was or whether it could be rolled over. As a result, the robot would either blindly crash into it or tragically get stuck in a corner.
Simply put, 2D LiDAR provides a flat, black-and-white photo, whereas robots need a 3D, full-color, dynamic map.
This is the primary background for the emergence of 3D LiDAR—robots cannot just know "whether something is blocking the way"; they must know "what it looks like, how high it is, how large it is, and how to bypass it."
II. Why Is 3D LiDAR Indispensable?
The principle is actually quite straightforward.
Its working principle is as follows: the device actively emits laser beams, which reflect back after hitting surrounding objects. By calculating the time of flight for the lasers to travel back and forth, the device accurately determines the distance and position of each object, ultimately stitching together a 3D representation of the entire environment, commonly known in the industry as a point cloud.
Currently, the mainstream solid-state LiDAR in the industry has completely eliminated the motors and rotating parts of traditional devices. It relies on self-developed chips to perform scanning, resulting in a simpler structure, longer lifespan, and higher stability.
Its "reliability" is mainly manifested in the following aspects:
First, all-weather stable operation, unaffected by environmental conditions. Ordinary cameras are highly dependent on light; day and night, cloudy or sunny, and backlighting all affect their performance. In contrast, LiDAR uses active light emission for detection. Whether in pitch-black nights, bright daylight, or rainy, snowy, and foggy weather, it can stably perceive the environment without going "blind."
Second, precise and error-free data. The depth distance of cameras relies entirely on algorithmic estimation, which is prone to errors. LiDAR, however, directly outputs centimeter-level precise 3D coordinates. It can accurately identify subtle obstacles such as steps, cables, and small debris, providing absolutely reliable data support for device navigation and obstacle avoidance.
Third, wide detection range and broad scenario adaptability. Compared to the detection range of a few meters for depth cameras, LiDAR can detect distances of tens of meters, which is sufficient to support long-range predictive obstacle avoidance and large-scale mapping and navigation for outdoor robots.
Fourth, increasingly prominent cost-effectiveness advantages. Leading manufacturers represented by RoboSense have achieved full-stack in-house R&D for chips, hardware, and mass production. A single technological foundation can be applied to both automotive and robotics products simultaneously. Through large-scale mass production, R&D and production costs are amortized, bringing the price of LiDAR down from the tens of thousands of RMB in the early days to the current affordable price of thousands of RMB, making it viable for large-scale commercial deployment.
III. The "Thrills" and "Pains" of Deploying in Humanoid Robots
Humanoid robots differ from ordinary robot vacuums and lawnmowers. They need to walk, climb up and down stairs, avoid obstacles, and operate autonomously in non-standard, complex scenarios such as homes, factories, and parks, just like humans. This places extremely high demands on the perception system, and 3D LiDAR is the core hardware for achieving autonomous operation.
First, it solves the "poor sense of direction" problem. The activity scenarios for humanoid robots change constantly, making it difficult for traditional perception solutions to achieve accurate mapping and localization. 3D LiDAR scans the surrounding environment in real time, quickly builds high-precision 3D maps, and corrects positions in real time during movement, avoiding localization drift and getting lost.
Second, it eliminates the challenge of safe obstacle avoidance in complex scenarios. Falls and collisions of humanoid robots are the biggest pain points for commercial deployment. Ground cables, low steps, scattered debris, and moving pedestrians are all safety hazards. LiDAR can identify obstacles over ten meters in advance and quickly adjust the walking trajectory regardless of day or night, making up for the shortcomings of pure vision solutions that are prone to missed detections and misjudgments.
Finally, it adapts to the need for lightweight body design. The internal space of humanoid robots is extremely limited. Traditional LiDARs are large and bulky; installing them on a humanoid robot would increase weight and affect battery life. Therefore, there is a strong need for a compact and lightweight 3D LiDAR product that can be easily integrated into the robot's head or torso without affecting the body structure and battery performance, to suit consumer-grade and industrial-grade humanoid robots. RoboSense's solid-state LiDAR perfectly meets this "want it all" requirement.
However, as the most core hardware in the vision system of humanoid robots, 3D LiDAR also has its shortcomings.
First, it has a "presbyopia" issue; the closer the distance, the more severe the perception blind spots. LiDAR cannot identify subtle objects within 0.3 to 0.5 meters. It is completely incapable of handling close-range precise operations such as grasping parts or water cups. It must be paired with depth cameras and structured light devices for auxiliary work, as a single LiDAR cannot cover all scenario requirements.
Second, performance degradation in extreme weather. Although its performance is a qualitative leap compared to cameras, in extreme weather such as heavy rain, dense fog, and dust storms, the laser beams will be scattered, significantly shortening the detection distance. The operational stability in extreme outdoor environments will decrease, requiring complementary use with millimeter-wave radar.
Third, limited recognition capability for special objects. Objects such as black fabrics, dark rubber, and matte furniture have high laser absorption rates, which will significantly shorten the detection distance. To improve recognition effects, device power consumption must be increased, creating a certain contradiction.
Fourth, hardware costs remain relatively high. Compared to ordinary depth cameras, the chip and module costs of solid-state LiDAR are higher, which will increase the overall price of humanoid robots in the short term. To achieve widespread adoption, continuous cost reduction through larger-scale mass production is still needed.
IV. How Will the Vision System of Humanoid Robots Evolve in the Future?
Here are a few interesting directions:
Chip integration, making the radar "smaller and smarter."
Just as smartphone chips become increasingly powerful, LiDAR is also taking a "digitalization" path. For example, the solid-state solution being explored by RoboSense has no mechanical rotating parts, resulting in a smaller size and lower cost. In the future, it might even be embedded into the robot's body just like a smartphone camera.
Sensor integration, no longer working in isolation.
Currently, the "head" of a humanoid robot is packed with a bunch of sensors: LiDAR, cameras, and IMUs, each calculating independently and then pieced together at the end. In the future, it may move towards hardware-level fusion, integrating these sensing components into a single integrated chip. With underlying time synchronization and spatial alignment, the data will naturally achieve "granularity alignment," making the robot react faster and less prone to processing conflicts.
Perception range extending from "how to move" to "how to work."
Currently, LiDAR mainly handles navigation (the job of the lower body). In the future, there may be a spin-off of close-range, high-precision radars specifically for fine manipulation (the job of the upper body), dedicated to grabbing screws, twisting parts, and performing assembly. This means humanoid robots will not only be able to walk but also become more adept at fine manual tasks.
Fourth, a "plunge" in prices, making it affordable for every robot.
According to industry forecasts, the market for 3D LiDAR used in robots could surge to over 4 billion RMB by 2030. As demand increases, costs will naturally drop. Once the price falls to the level of a few hundred RMB, the installation rate in humanoid robots will shift from an "optional extra" to a "standard configuration."
From "wall-bumping exploration" to "3D panoramic vision," 3D LiDAR enables humanoid robots to clearly see the outline of the world. Although it is not omnipotent at present, without it, the safety and autonomy of humanoid robots would be significantly compromised. It can be said that on the path to truly "perceiving like a human," 3D LiDAR is not the endpoint, but it is absolutely the bridge that cannot be bypassed.
Who will dominate the future, and which will be superior among cameras, lasers, and millimeter-wave radars? The high probability is that none can do without the others. Only by selectively integrating various technologies and gradually making up for visual shortcomings can humanoid robots be nurtured into true masters capable of "seeing in all directions" and reacting swiftly.
Data sourced from publicly available information on the internet. Corrections and additions are welcome. This article represents only the author's views and does not constitute any investment advice.