August 5, 2020, Suzhou, China (Baidu Apollo autonomous driving computing platform ACU mass production roll-off ceremony) — Xilinx Corporation (NASDAQ: XLNX), the global leader in adaptive and intelligent computing, today announced that Baidu’s Apollo autonomous driving computing platform ACU (Apollo Computing Unit) equipped with Xilinx’s automotive-grade chip platform Zynq UltraScale+ MPSoC was officially mass-produced at Flextronics’ Suzhou plant in China. This mass-produced ACU hardware platform will be the first to be applied to Apollo Valet Parking (AVP) autonomous parking products. More than 30 representatives from world-renowned OEMs and parts suppliers gathered in Suzhou to witness the glorious moment of the industry’s first mass-produced autonomous driving computing platform from the laboratory to the production line. At the same time, Baidu also said that it has reached cooperation agreements with a number of OEMs. Among them, WM Motor, a new force in Chinese car manufacturing, will be the first to carry Baidu AVP on board this year.
Figure 1: Tang Xiaolei (third from left), vice president of sales for Xilinx Greater China, and Wang Yunpeng (fourth from right), general manager of Baidu’s Autonomous Driving Technology Department, congratulate Baidu’s ACU platform for mass production
Baidu ACU is Apollo’s mass-produced autonomous driving in-vehicle computing unit. It is divided into multiple series of products according to the computing power requirements of different demand scenarios. At present, the annual production capacity of the ACU mass production line can reach 200,000 sets. Among them, ACU-Advanced is the industry’s first dedicated in-vehicle computing platform for autonomous parking products AVP, providing a solution that integrates software and hardware. The traditional automatic parking function mainly relies on ultrasonic radar, and autonomous parking needs to be safer in environmental perception. Therefore, more complex sensor fusion needs to be combined with the parking system, and it also requires advanced and powerful deep learning inference capabilities. and more powerful computing power to handle complex driving environments. Leveraging the powerful sensor fusion and AI processing capabilities of the Xilinx automotive-grade heterogeneous multi-core Zynq UltraScale+ 5EV platform, the new platform can support up to 5 cameras and 12 ultrasonic radars. In addition, millimeter-wave radar and lidar interfaces are reserved, and are adapted to the Baidu PaddlePaddle deep learning framework to support rapid algorithm iteration and upgrade.
Figure 2: Baidu Apollo Autonomous Driving Computing Platform
Wang Yunpeng, general manager of Baidu’s Autonomous Driving Technology Department, said: “Baidu has been continuously cultivating and accumulating in the field of autonomous driving, and the mass production of ACU indicates that we are in vehicle-grade functional safety, commercialization of autonomous driving products, and software and hardware supply chain management. Significant enhancement of capabilities. We work with top partners in the industry to work together on ACU R&D and design. As the world’s leading FPGA supplier, Xilinx provides core processing chips for Baidu ACU, and has a vehicle-grade level that other solutions cannot provide. production performance.”
Figure 3: ACU from lab to production line
Tang Xiaolei, Vice President of Xilinx Greater China Sales, attended the mass production roll-off ceremony of Baidu’s ACU platform and delivered a keynote speech on “Building a High-Performance and High-Reliability AVP System”. She said: “We are very pleased to see Xilinx help Baidu successfully mass-produce the industry’s first ACU autonomous driving computing platform. Baidu is rooted in China and has gradually become a global leader in the field of autonomous driving. We are proud to provide drivers with a safe and convenient smart driving experience.”
Xilinx has nearly 20 years of industry experience in the automotive business, and has consistently maintained double-percent annual growth in shipments for the past 15 years. In particular, the success of the 28nm and 16nm product lines has become a catalyst for the growth of Xilinx’s automotive business. As of today, Xilinx has sold more than 190 million automotive-grade chips worldwide, of which more than 75 million are used in mass-produced ADAS (Advanced Driver Assistance Systems). Currently, Xilinx serves more than 200 automotive companies, including major global Tier 1 automotive suppliers, equipment manufacturers (OEMs) and various start-ups.