China's Baidu Announces That the World's First Autonomous Driving Compute Unit is Ready for Production

author: Eric Walz   

The race to develop self-driving vehicles is not just in Silicon Valley. In China, the world's biggest auto market, companies are also rapidly developing autonomous driving technology, including China's internet search giant Baidu, which is often referred to as the "Google of China."

For the past three years, Baidu has been developing autonomous driving technology with over 100 industry partners as part of its open "Apollo" project. Now the company says the world's first production-ready compute platform specifically for autonomous vehicles is ready for application.

Baidu calls it the "Apollo Computing Unit" (APU). The ACU will support Baidu's autonomous parking feature called Apollo Valet Parking (AVP). AVP was jointly developed by Baidu and Chinese electric vehicle maker WM Motor. The automaker will be the first automaker offering the valet parking feature at scale beginning in the second half of 2020.

All of the ACU's components meet automotive-grade production requirements. Baidu said the production capacity of the ACU is 200,000 units a year.

Autonomous parking capability is a popular vehicle feature for Chinese car buyers, as its very difficult to park in many of China's crowded cities.

Apollo Valet Parking can enable vehicles to perform remote pick-up or return and valet parking using a mobile device and vehicle app. It's similar to Tesla's Summon feature where the vehicle drives to the owner's location in a parking lot, or parks itself after the driver exits.

Baidu Partnered the industry's Leading Suppliers on R&D and Design of the Apollo Computing Unit

The production-ready ACU features a field programmable gate array (FPGA) from Silicon Valley semiconductor manufacturer Xilinx that's used for sensor fusion and AI processing. The platform is fully compatible with Baidu's open source "PaddlePaddle" deep-learning framework.

Baidu open sourced PaddlePaddle in 2016 in order to further support innovation. It's a comprehensive and powerful deep learning platform designed for developers. PaddlePaddle supports innovation by providing developers with the tools and resources they need to rapidly adopt and implement their deep learning projects at scale.

PaddlePaddle is currently used by more than 1.5 million developers.

For the task of autonomous parking, PaddlePaddle supports five vehicle cameras and 12 ultrasonic radars.

In the ACU project with Baidu, Singapore-based electronics manufacturer Flex is providing hardware development, diagnostic software development, thermal management design, test verification, as well as large-scale manufacturing. Baidu first signed a strategic partnership agreement with Flex in 2018 to develop hardware solutions for autonomous driving, vehicle connectivity and smart traffic systems.

"The mass production of ACU indicates that we have significantly advanced our capabilities in automotive-grade safety, commercialization of autonomous driving, and software and hardware supply chain management," says Yunpeng Wang, senior director of the Technology Department of Baidu's Intelligent Driving Group. He added that "Baidu and Flex have formed an effective technology alliance by sharing advantageous resources and complementing technologies, and achieved a high degree of integration of intelligence and industrialization."

For the ACU's microcontrollers, Baidu Apollo partnered with leading semiconductor Infineon to manufacture the automotive-grade components.

Baidu also worked with Vector for the ACU project. The company is a leading AUTOSAR (Automotive Open System Architecture) system supplier, to provide the underlying software operating system for ACU. Vector provides the auto industry with a professional platform of tools and software components for the development of embedded systems.

Baidu's Apollo Platform

Baidu Apollo is an open source platform that's designed to speed up the global development of autonomous driving technologies through close collaboration between industry partners. Apollo is a comprehensive, secure, all-in-one, complete autonomous driving ecosystem which supports all major functions of a self-driving car. Baidu's open platform is regarded by many as the "Android" of the automotive industry. It was first announced in April 2017. 

Apollo includes cloud-based access to high-definition maps, perception, path planning and a driver simulation engine, which allows testing of autonomous driving behaviors without endangering the public. The Baidu simulation engine has embedded HD maps with centimeter level accuracy and is backed by vast amounts of real-world driving data.

In March 2018, Baidu announced the release of "Apollo Scape", billed as the world's largest open-source dataset for autonomous driving technology. The dataset is available for developers as part of Apollo. It eliminates the time it takes for manual data collection. Datasets are typically used in machine learning for regression and classification tasks.

In May, Baidu announced the completion of its "Apollo Park" autonomous driving test facility, which the company says is the world's largest intelligent vehicle infrastructure cooperative system (IVICS). 

Apollo Park can support all of the functions associated with self-driving and connected vehicles and will support the further development of autonomous vehicles, as well as vehicle-to-vehicle (V2V) and vehicle-to-everything (V2X) technology, which allows vehicles to communicate with each other as well as to urban infrastructure, such as traffic lights. 

In addition to developing self-driving technology and autonomous parking, Baidu is developing a next-generation intelligent in-car interactive operating system that's integrated with Baidu's "DuerOS" robust conversational AI platform. It allows drivers to interact with most vehicle systems using only their voice.

Baidu Apollo automotive partners include BMW, Daimler, Ford Motor Co, Toyota, Volkswagen and Honda, as well as chipmakers Intel and Nvidia.

Nvidia is also making powerful AI-based processors for future autonomous vehicles. Like Apollo, the company's NVIDIA DRIVE autonomous vehicle platform is open to developers worldwide. It includes DRIVE software plug-ins that allow Nvidia's automotive partners to integrate their self-driving technology with the company's DRIVE platform.

Chinese electric vehicle startup Xpeng Motors is the first car company in China to install Nvidia hardware during production. The automaker's P7 smart sedan on sale since April is powered by the new DRIVE AGX platform, which features Nvidia's high-performance Xavier processor. 

Eric Walz

Originally hailing from New Jersey, Eric is a automotive & technology reporter covering the high-tech industry here in Silicon Valley. He has over 15 years of automotive experience and a bachelors degree in computer science. These skills, combined with technical writing and news reporting, allows him to fully understand and identify new and innovative technologies in the auto industry and beyond. He has worked at Uber on self-driving cars and as a technical writer, helping people to understand and work with technology.