![]() |
市场调查报告书
商品编码
2007697
光达(LiDAR)的应用现况(汽车、通用机器人等)(2025-2026 年)LiDAR (Automotive, Pan-Robotics, etc.) Application Research Report, 2025-2026 |
||||||
概括
预计到 2025 年,中国乘用车将配备 321 万个雷射雷达感测器,比上一年增长 110.1%。
高通道资料的数位化代表着从「点云」到「影像」的世代飞跃。
随着竞争加剧,多感测解决方案正逐渐成为主流。
从汽车应用到通用机器人应用的扩展正在加速。
2025年,比亚迪、长安汽车、理想汽车等主要汽车製造商持续推行「技术均等化」策略。因此,先进的自动驾驶功能不再是高阶车型的专属,而是逐渐普及到所有车型。在此背景下,作为城市自动驾驶等先进功能核心感测器之一的光达(LiDAR)的搭载率不断提高。为了支援城市自动驾驶,包括2026年领克08 EM-P、威高山7、爱拓M6和东风亿京在内的多款车型均标配雷射雷达。
在中国,强制国家标准GB 39901-2025《小型车辆自动紧急煞车系统技术要求及试验方法》将于2028年1月1日生效。届时,所有新车必须符合该标准的要求,而已获得型式认可的车型将有12至24个月的过渡期以完成技术合规。这项新的国家标准对自动紧急煞车系统的性能提出了更严格的要求,尤其是在夜间等复杂条件下的性能。为了满足未来日益严格的自动紧急煞车系统测试标准,光达已成为不可或缺的感测硬件,这体现了政策对雷射雷达长期搭载率的预期。
随着车载自动驾驶系统的普及,到 2025 年,中国乘用车上将安装超过 500 万个雷射雷达感测器。 2025 年,中国乘用车上安装的雷射雷达感测器数量达到 321 万个,比上一年增长 110.1%。
根据过去三年的月度数据,中国乘用车雷射雷达(LiDAR)的搭载率持续稳定上升。 2025年5月,得益于力狮L6和2025款爱拓M9等热门车款的上市,搭载率突破10%。 2025年11月,搭载率搭载率18.5%,创历史新高。
近年来,光达的「通道数」(垂直解析度)已成为衡量其性能的重要指标。主流产品正迅速从128通道和192通道型号过渡到更高规格型号。例如,超高通道数雷射雷达包括华为的896通道双光路影像级雷射雷达、合赛的ETX(800通道)和AT1440雷射雷达(1440通道),以及RoboSense的EM4,其通道数可从520通道定製到2160通道不等。预计到2026年,500通道光达将成为旗舰智慧汽车车型的“新标准”,而192通道光达有望成为大众汽车20万元以下车型的标配。
本报告对光达市场进行了深入分析,提供了有关光达概述和发展趋势、技术结构和组件、主要元件供应商、国内外领先的光达供应商、光达技术比较以及光达在各个领域(汽车、通用机器人)的应用等资讯。
定义
LiDAR research: hardware competition shifts to combined sensing capabilities from "point cloud" to "images" and from automotive to robots
The "LiDAR (Automotive, Pan-Robotics, etc.) Application Research Report, 2025-2026" mainly covers the following content: LiDAR overview and development trends, technical structure and components, market data analysis, core component supplier research, domestic and foreign top LiDAR providers, LiDAR parameter summary, LiDAR technology comparison, LiDAR application in different fields (automotive, pan-robotics), etc.
【Abstract】
In 2025, China's passenger cars installed 3.21 million LiDAR sensors, a year-on-year increase of 110.1%.
High-channel digitalization brings an intergenerational breakthrough from "point cloud" to "images"
The competition intensifies and combined sensing solutions gradually become mainstream
The expansion from automotive to pan-robotic applications is accelerating
In 2025, leading OEMs represented by BYD, Changan, Li Auto, etc. continued to promote the "technological equality" strategy, resulting in high-level autonomous driving functions no longer being exclusive to top-of-the-channel vehicle models, but becoming common to all vehicle models. In this context, LiDAR, as one of the core sensors for realizing high-levl functions such as urban NOA, has seen its installation rate rise. In order to support urban NOA, many vehicle models have LiDAR as standard, such as the 2026 Lynk & Co 08 EM-P, WEY Gaoshan (Alpine) 7, AITO M6, Dongfeng Yijing, etc.
China will begin to implement the mandatory national standard "Technical Requirements and Test Methods for Light Vehicle AEB Systems" with the standard number GB 39901-2025 from January 1, 2028. By then, all new vehicle models must meet the requirements of the standard, and vehicle models that have already received type approval will have a transition period of 12 to 24 months to complete technical adaptation. The new national standard has stricter requirements for AEB performance, especially under complex scenarios such as at night. To meet the more stringent AEB test standards in the future, LiDAR has become an indispensable sensing hardware, which provides policy expectations for the long-term installation rate of LiDAR.
With the promotion of autonomous driving systems in vehicles, China's passenger cars had installed over 5 million LiDAR sensors by 2025. In 2025, China's passenger cars installed 3.21 million LiDAR sensors, a year-on-year increase of 110.1%.
As per the monthly data in the past three years, China's passenger car LiDAR installation rate has maintained a steady upward trend. In May 2025, the LiDAR installation rate exceeded 10%, mainly driven by popular vehicle models such as the newly launched Li L6 and the 2025 AITO M9. In November 2025, the LiDAR installation rate hit 16.6%, which was mainly affected by the robust growth of new vehicle models such as Xiaomi YU7 and 2026 AITO M7. In December 2025, LiDAR set a new high, with the installation rate reaching 18.5%, thanks to Fangchengbao Titanium 7, 2026 NIO ES8, etc.
In recent years, LiDAR's "channel count" (vertical resolution) has become a key indicator of its performance. Mainstream products have rapidly moved from 128 channels and 192 channels to higher specifications. For example, LiDAR with ultra-high channel count includes Huawei's 896-channel dual-optical path image-level LiDAR, Hesai ETX (800 channels) and AT1440 LiDAR (1440 channels), and RoboSense EM4 supports customization from 520 channels to 2160 channels. In 2026, 500-channel LiDAR has become the "new threshold" for flagship smart vehicle models, while 192-channel LiDAR will spread to Volkswagen's vehicle models priced below RMB200,000 as standard.
On March 4, 2026, Huawei Qiankun released the next-generation dual-optical path image-level LiDAR, which pushed the mass production of 896-channel LiDAR, marking the current highest level in the world. It was first seen in the new versions of Maextro S800 and AITO M9, handling long-tail scenarios of autonomous driving (corner cases):
The recognition distance is significantly extended: the recognition distance of 896-channel LiDAR for low-reflectivity objects (such as black tires) is increased from 42 meters to 122 meters, and the recognition distance of special-shaped obstacles (such as cones) is increased by 77%. This provides longer decision-making reaction time for high-speed driving and fundamentally improves active safety capabilities.
The recognition accuracy is higher across generations: the minimum object height that can be stably recognized has been reduced from 30 cm to 14 cm, and small obstacles such as small cartons, gravel, and fallen traffic cones on the pavement, which were easily ignored in the past, can be accurately recognized. In a night scenario, it can even clearly see the details of a puppy wagging its tail 55 meters away, reaching an "image-level" perception level.
High channel count directly brings exponential improvement in angular resolution. The vertical angular resolution of 192-channel radar is usually around 0.2°, while radar with more than 500 channels can compress the angular resolution to 0.05° or even 0.01°, (such as AT1440 with 0.05° x 0.0125°, AT128 with optimal 0.1° x 0.2°, ATL with 0.08° x 0.1°, EMX with 0.08° x 0.10° global resolution, EM4 with 0.050°X0.025°, Falcon K3 with up to 0.07°X0.03°).
Hesai Technology's AT1440 is an automotive-grade ultra-high-definition LiDAR for high-level autonomous driving. It features 1440 channels and image-level point cloud output, with a single echo point frequency of up to 34 million points/second. It is equipped with Hesai's fourth-generation self-developed chip, with cutting-edge efficient photosensitivity and ultra-high parallel processing technology. The detection range is up to 300 meters @10%, mainly for L4 and above autonomous driving systems. It is the core sensor of the "Qianliyan A" sensing solution. This solution usually uses four AT1440 sensors to work together to achieve 360° full coverage and zero blind zone sensing. Applications include robotaxis, robotrucks, etc.
Seyond's Falcon K3 long-range LiDAR has an equivalent of 600 channels, an angular resolution of up to 0.07° X 0.03°, and a detection range of 350 meters. It is mainly mounted on NIO's high-end vehicle models and provides key sensing support for L3 and higher-level autonomous driving. 600,000 units had been delivered as of early January 2026 cumulatively.
Behind the resolution jump is the paradigm shift from analog architectures to digital chips, and the evolution from analog signal (APD) architectures to digital signal (SPAD-SoC) architectures. The SPAD (Single-Photon Avalanche Diode) chip is homologous and isomorphic to the camera CMOS, and uses a pixel array, so that increasing the number of channels is like increasing the number of pixels in a camera, and is no longer subject to the physical limitations of the number of analog channels and complex circuits.
For example, RoboSense's EM4 uses VCSEL + SPAD-SoC and integrates many advanced technologies such as digital architecture, crosstalk elimination, full-condition photoelectric signal processing, and lossless data compression. Based on the platform design, EM4 supports customized technical solutions such as 520 channels, 720 channels, 1080 channels, and 2160 channels. At present, the custom-developed 520-channel version, relying on its mature and mass-produced advantages, has been mass-produced for many vehicle models such as IMLS9, IMLS6, and ZEEKR 9X.
EM4 can provide 1080P high-definition three-dimensional perception for vehicles, with the imaging of 25.92 million points/second. It can not only measure distances up to 600 meters, but also accurately recognize distant small objects such as tires, cones, and cartons. Compared with current mainstream LiDAR sensors, EM4 can increase the response time of the autonomous driving system by up to 70%, making the system decision-making and response calmer, and making the autonomous driving experience safer and more comfortable.
The high channel count of automotive LiDAR expands the safety boundary of autonomous driving from "high-probability visible" to "very small probability visible details", providing better "safety redundancy." This transition is the key cornerstone for high-level autonomous driving at L3 and above to move from "available" to "reliable".
Currently, global autonomous driving is in a critical transition from L2+ to L3. At the same time, the overall sales volume of Chinese automobiles is under pressure in 2026. LiDAR, as one of the core components of autonomous driving perception, has upgraded its industry competition from competing on radar performance to providing an overall solution capability of "hardware combination + algorithm collaboration + scenario adaptation".
Regarding the safety standards of different autonomous driving levels, Hesai is an example. In the L2 market, it mainly uses ATX (costing about US$200) to promote the transformation of LiDAR from "high-end configuration" to "safety standard configuration"; in the L3/L4 market, a higher safety limit is defined through high-performance combinations, such as the L3 LiDAR combination (ETX*1 + FTX*2), and the L4 LiDAR combination (AT1440*4 + FTX*4).
Other combined sensing solutions include:
Seyond's "1+2 LiDAR combination" solution: 1*Falcon ultra-long-range main-view LiDAR + 2*Robin W wide-angle LiDAR. These three LiDAR sensors are integrated into NIO's new ES8, building an all-scenario perception matrix from long range to near field, from main view to blind filling.
VanJee Technology's WLR-760 + WLR-750 autonomous driving perception system: Currently available on ZELOSZ5, its perception system consists of two WLR-760s forward LiDAR sensors and two WLR-750 side and rear LiDAR sensors, achieving all-round, high-precision environmental perception. Z5 is geared towards urban logistics and distribution, with flexible container configuration and multi-mode distribution.
MicroVison's Tri-LiDAR architecture (1*MAVIN + 2*MOVIA S): With the Tri-Lidar architecture, MicroVision provides multiple LiDAR sensors (such as two short-range (MOVIA S) LiDAR sensors and one long-range (MAVIN) LiDAR sensor) integrated into a unified open platform design. This design allows long-range radar (such as the simplified Mavin) to focus on farther detection, while corner LiDAR covers the edge of the field of view, so that they work together to achieve comprehensive environmental perception.
RoboSense's EM4+E1 combination has become one of the preferred sensing solutions for the next-generation robotaxi, and has completed production verification with 8 leading customers around the world. In other fields, for example, on September 15, 2025, RoboSense and MINIEYE reached strategic cooperation. MINIEYE's next-generation iRobo autonomous vehicles - T5 and T8 will be equipped with 3 digital LiDAR sensors from RoboSense, including an automotive-grade 192-channel EMX LiDAR sensor and 2 E1R all-solid-state blind-filling LiDAR sensors, which can greatly improve the perception accuracy and safety redundancy under complex traffic scenarios.
In addition to automotive applications, LiDAR is being implemented on a large scale in various robot scenarios, providing precise sensing capabilities for AGVs, AMRs, and quadruped and humanoid robots. Among them, lawn mowing robots and autunomous delivery robots have become the main application scenarios. Currently, representative products include Livox's MID-360, Lanhai Photoelectricity's LDS-M300, RoboSense's Airy/E1R, Hesai's JT series, etc.
For example, RoboSense sold 303,000 LiDAR sensors in the robotics field in 2025, which are widely used in scenarios such as intelligent lawn mowing robots, autunomous delivery, and humanoid robots. Compared with 2024, RoboSense's sales volume in the robotics field grew strongly by 1141.8% year-on-year.
Since its release in January 2025, Hesai's JT 3D LiDAR has experienced explosive growth in the robotics market thanks to its outstanding performance. From its release to May 2025, Hesai quickly achieved the milestone of delivering 100,000 units; by the end of 2025, it had cumulatively delivered more than 200,000 units. The robot application scenarios include agriculture (Agtonomy), cleaning (Gausium), lawn mowing (Dreame), airports (Boenic), smart measurement (Realsee), game modeling (Black Myth: Wukong), and factory automatic scheduling (BMW's factory in Germany). In the field of logistics robots, Hesai is also assisting Meituan UAVs, ZELOS autonomous vehicle, Neolix, etc. to bring innovative solutions to urban distribution & logistics.
By expanding into "pan-robotics", leading LiDAR companies can not only multiplex their platform technologies to the robotics arena to support the dual growth of "ADAS + robots", but also diversify business risks and find new growth engines.
Definitions