Bitauto News Recently, Hesai officially announced the release of the long-range hybrid solid-state lidar AT128 for ADAS front-mounted mass production, helping the rapid realization and popularization of L3 autonomous driving. AT128 is the only automotive-grade front-mounted mass-produced lidar on the market that meets the requirements of both long-range (200m@10) and ultra-high point frequency (1.53 million per second, single echo). At present, AT128 has received orders from a number of top automobile manufacturers for a total of more than 1.5 million units, and will be mass-produced and delivered in 2022.
AT128 has achieved the ultimate in several important core indicators, which will greatly improve perception capabilities and reduce the difficulty of algorithm adaptation in usage scenarios.
Image-level resolution, ultra-high pixel frequency of 1.53 million per second
Resolution determines the object details that a lidar can effectively capture. Compared with the technology on the market that achieves "equivalent XX lines" through two-dimensional scanning and splicing, Hesai AT128 achieves "true 128-line" structured scanning through chip-based 128-channel solid-state electronic scanning, avoiding the need for two-dimensional high-speed While mechanical scanning has an impact on product reliability and lifespan, it also achieves a complete and evenly distributed point cloud in the horizontal and vertical directions without splicing. The structured data similar to a camera can bring great convenience to post-processing algorithms. .
There are many opportunities for trickery in the lidar market’s description of resolution. For example, some descriptions of "up to 0.0X degree resolution" do not specify the refresh frame rate. We believe that the 0.02° resolution at a 1Hz frame rate has no reference significance for autonomous driving. Similarly, high resolution achieved only within a small angle range, or local encryption achieved by dynamically adjusting the ROI, cannot comprehensively improve the system's perception capabilities. Instead, it may cause more inaccuracies due to inaccurate ROI selection. Certainty.
Point frequency is the most direct reflection of lidar resolution and comprehensive imaging capabilities. It is similar to the concept of total pixels of a camera. It can neither rely on lowering the refresh rate to make tricks, but also avoids the size and unevenness of the field of view. Bringing such a one-sided indicator as "highest resolution".
Super range-measuring capability helps high-end autonomous driving
Equally important as the point frequency is the range-measuring capability of lidar. After looking at things in the fog in the past four or five years, mainstream players and customers in the industry have reached a standard definition of distance measurement capability, which is the stable measurement distance of objects with a reflectivity of 10. The ADAS industry generally believes that 200m@10 is a necessary condition for achieving Level 3 functions stably and reliably.
Hesai AT128 not only has a ranging capability of 200 meters @ 10, but the furthest ground line can reach 70 meters. Due to the limitation of the galvanometer caliber, the effective ground detection distance of most MEMS systems generally does not exceed 40 meters, and the total sensing distance will be greatly reduced accordingly.
Since distance measurement is so important, would it be better if the 905nm light source was replaced with a 1550nm light source and the distance could be increased to 250m@10? uncertain. Due to the power and heat limitations of the 1550nm device, its point frequency drops significantly at the same power and heat level. Currently, no company can achieve a point frequency of more than 1.53 million points per second like the AT128. In a typical 1550 solution, the measurement range is 25 farther, but the total points are 50 less. The final result is that the number of effective points used for the perception algorithm at a distance does not increase but decreases, which is counterproductive in the end.
Picture: Ranging vs. point frequency comparison of mainstream lidar solutions on the market
Automotive grade is produced for front-mounted mass production
As a mass production oriented to front-mounted The AT128 is a manufactured lidar that is positioned as a vehicle-grade ultra-high reliability radar and complies with the full-process closed-loop vehicle-grade requirements from design to verification. The core components of AT128 adopt vehicle-grade devices (meeting AEC-Q and other relevant standards). The DV test covers more than 50 vehicle-grade tests in accordance with international OEM standards such as electrical, mechanical, climate, sealing, materials, and electromagnetic compatibility.
AT128 functional safety solution covers detailed FMEA/FTA analysis of optical, mechanical, electrical and soft subsystems, taking into account thousands of different failure modes and response strategies. Hesai has designed complete fault diagnosis functions for AT128 through complete failure analysis and fault injection. The fault coverage rate is as high as 90% or more, and the fault response time is less than or equal to 100ms. At the same time, it is equipped with an active anti-interference function that is valued in mass production projects, and is equipped with a diagnostic function that is accurate to each line. AT128's SOP mass-produced products meet the requirements of functional safety ASIL-B(D).
AT128 has a comprehensive network security protection system. In accordance with the ISO21434 development system, Hesai uses data signature and encryption, client authentication and other methods to achieve safe upgrades, safe startup, and prevent hackers from intruding on the lidar for tampering, counterfeiting and other damage.
In order to support the rapidly growing mass production orders, Hesai is building a Hesai "Maxwell" intelligent manufacturing center with a construction area of ??over 60,000 square meters and a designed annual production capacity of over one million units, and is expected to open in 2022 Fully put into production.
The intelligent manufacturing center will undertake functional module tasks such as lidar R&D, production, and vehicle specification testing, and introduce advanced "automated flexible" production lines to achieve flexible production capacity and multiple demand adaptation capabilities, and gradually become A super R&D and manufacturing center with strong delivery capabilities can meet the diversified and customized mass production needs of customers to the greatest extent.