At this Beijing Auto Show, there were actually not many new cars unveiled, but several Tier 1 players in the autonomous driving industry were making frequent moves. We also visited several key manufacturers in the industry and gained some understanding of their recent trends. Today we will share them with you.
On September 28, at the parts exhibition area of ??the Beijing Auto Show, Black Sesame launched a fully autonomous driving computing platform called FAD (Full Autonomous Driving). Its performance directly benchmarked NVIDIA Xavier and Tetra SLA FSD platform.
The FAD fully autonomous driving computing platform released by Black Sesame Intelligent Technology is based on the cascade solution of two Huashan No. 2 A1000 chips. The computing power can reach 80TOPS-140TOPS, and the overall energy efficiency ratio is as high as ?6TOPS/ W.
Officially, the dual-chip FAD platform has sensing, positioning, sensor fusion, path planning and vehicle control functions, and can meet the needs of TJP (Traffic Jam Pilot), HWP (High Way Pilot) ), CP (Car Park Pilot) and other complete L2/L3 intelligent driving scenarios.
Judging from the products on display, the circuit board area of ??the platform is relatively large, the two Huashan A1000 chips are very eye-catching, and the heat sink area is also very large. It was learned from the on-site staff that the platform can achieve passive heat dissipation due to its high energy efficiency, and the fanless design can also improve system stability to a certain extent.
At the same time, we also learned that this platform was developed in cooperation with FAW Group, so FAW Group products such as Hongqi may be the first to adopt this platform. At the same time, other brands are also negotiating with Black Sesame.
It is reported that in the future, Black Sesame will also provide a four-chip combination solution with a computing power of 280TOPS, which can support L4 level autonomous driving scenarios.
On the same day, Leishen Intelligent released the automotive-grade CH series hybrid solid-state lidar CH120 and CH64 at the parts exhibition hall. They have 120 lines and 64 lines respectively to flexibly meet the specific needs of different autonomous driving subdivisions such as field of view and resolution.
CH120 is the latest product of Leishen CH series. It is a 120-line hybrid solid-state laser radar with a horizontal field of view of 60°, a vertical field of view of -13°~7°, and a detection range of up to 250 meters, with an accuracy of ±2 cm, and a measuring point rate of up to 400,000 points per second.
CH120 is designed strictly in accordance with automotive grade standards. The overall weight is about 2.15kg. It can work in an environment of -40℃~85℃, and the power consumption is only 10w.
Compared with CH120, CH64 is smaller in size, lower in cost, weighs about 1.5kg, has a larger horizontal field of view of up to 120°, and a vertical field of view of -13.33°~7.67°. , can be flexibly embedded in various vehicles.
Currently, many domestic and foreign autonomous driving companies have obtained CH120 and CH64 in advance, and are conducting verification and testing on commercial vehicles such as mining trucks, heavy trucks, container trucks, logistics vehicles, buses, and minibuses. .
During the on-site communication with the staff, the author learned that Leishen Intelligent is currently in contact with a number of vehicle manufacturers. For example, the ideal next-generation new car is very likely to be equipped with Leishen Intelligent’s solid-state Radar, which also means that the ideal next-generation autonomous driving platform may use sensor fusion technology.
At the pre-show press conference on September 25, Huawei launched the MDC car-grade intelligent driving computing platform, Huawei Smart Car Cloud Service 2.0, Harmony OS smart cockpit, and all-in-one electric drive system Solutions such as DriveONE.
The MDC car-grade intelligent driving computing platform adopts a unified hardware π architecture, providing 48~160?TOPS of powerful computing power. It is based on the intelligent driving operating systems AOS, VOS and MDC?Core, and can cover L2 ~L4 level passenger cars, commercial vehicles, work vehicles and other intelligent driving application scenarios.
Among them, MDC?210 has a computing power of 48TOPS, mainly for L2 level autonomous driving, and MDC?610 can provide 160TOPS of computing power, for L3-L4 level autonomous driving. Both meet the requirements of vehicle-level safety Require.
It is reported that Huawei’s MDC solution integrates Huawei’s self-developed Host CPU chip, AI chip, and ISP chip, and is optimized through the integration of underlying software and hardware to achieve time synchronization and accurate sensor data. It leads the industry in processing, multi-node real-time communication, minimized noise floor, low power management, fast and secure startup, etc. The platform hardware follows mainstream interface standards such as sensors and actuators.
In addition, Huawei’s booth also displayed a variety of components independently developed by Huawei, such as millimeter-wave radar, lidar, binocular cameras, wide-angle lenses, etc.
At the Beijing Auto Show, Horizon brought us the iterative product of the Journey 2 - the Journey 3 domestic chip.
According to reports, Journey 3 uses a 16-nanometer process and is based on the BPU2.0 architecture independently developed by Horizon. It has an AI computing power of 5 TOPS and a typical power consumption of only 2.5W. It has high performance and low power. It features high power consumption, strong scalability, safety and reliability, and supports multiple application scenarios such as high-level assisted driving, smart cockpit, automatic parking assistance, high-level autonomous driving, and crowdsourced high-precision map positioning.
Official personnel said: "Jhengcheng 3 has extremely high AI computing power efficiency, energy consumption ratio surpasses many mainstream chips in the industry, and has excellent image access and processing capabilities. It not only supports deep learning-based It has functions such as image detection, classification, and pixel-level segmentation, and also supports efficient encoding of H.264 and H.265 video formats, making it an ideal platform for multi-channel AI computing and multi-channel digital video recording. ”
In addition, Horizon announced that it will soon launch the more powerful Zhengcheng 5 car-grade chip, which is oriented to high-level autonomous driving scenarios. The single chip has an AI computing power of 96 TOPS and supports 16 cameras. The composed autonomous driving computing platform has 192 -384?TOPS computing power, oriented to L3-L4 level autonomous driving, and meets the requirements of ASIL?D, the highest safety level in the automotive industry.
This article comes from the author of Autohome Chejiahao and does not represent the views and positions of Autohome.