On April 2, Baidu was invited to participate in the "2023 China Electric Vehicle Hundred People's Meeting - Digital Forum of the Automobile Industry". During the forum, Baidu officially released a new generation of autonomous driving cloud product - Apollo Cloud 2.0. Apollo Cloud 2.0 is a product that provides full-process cloud services for intelligent driving mass production for automotive enterprises. It has three core advantages: accurate compliance, multimodal large models, and city level simulation. It provides users with autonomous driving cloud services in mass production areas, helping automotive enterprises span industrial gaps and achieve rapid mass production.

　　Gao Guorong, Vice President of Baidu Intelligent Driving Business Group and General Manager of Intelligent Internet Connection Business

　　Gao Guorong, Vice President of Baidu's Intelligent Driving Business Group and General Manager of Intelligent Internet Connection Business, said, "Baidu is the earliest enterprise in China that has laid out the research and development of core technologies for autonomous driving. It has been paying close attention to the deep needs of automotive enterprises in the process of automotive intelligence. The launch of Apollo Cloud 2.0 will help automotive enterprise users overcome the key long tail problem and usher in the era of intelligent driving mass production."

　　The intelligent driving market is about to explode, and Apollo Cloud 2.0 is emerging as the times require

　　The new four modernizations of automobiles are rapidly evolving, with the development speed of electric vehicles as the representative in the first half far exceeding market expectations, accelerating the arrival of the intelligent era. According to relevant market reports, in 2021, the overall assisted driving insurance coverage of new passenger vehicles in the Chinese market exceeded 8 million, with a penetration rate of 40%. Among them, the carrying capacity of L2 class ADAS was 3.9 million vehicles, a year-on-year increase of 77%. The research report released by McKinsey also points out that intelligent driving has become one of the choice factors for consumers to make purchasing decisions.

　　According to Baidu's prediction, the penetration rate of high-end intelligent driving end users will exceed 15% in 2026, when smart cars will cross the critical point of development and usher in a big explosion. In the next three years, cars without autonomous driving capabilities may lose competitiveness. It is crucial for vehicle companies to plan ahead with mass production research and development, becoming the key to winning the intelligent driving market.

　　Currently, most of the autonomous driving cloud products in the industry mainly support the research and development of autonomous driving functions "from scratch", and cannot cope with the challenges of compliance, efficiency, experience, cost, and other long tail issues faced in the mass production stage, which determine the success or failure of intelligent driving mass production. Baidu has rich research and development and practical experience in the field of autonomous driving, and has significant advantages in tackling the problem of intelligent driving with long tail. The successful launch of Apollo Cloud 2.0 has achieved a milestone evolution from the R&D tool chain to intelligent production lines, building a vehicle intelligent development and production line for automotive companies, and providing full process cloud services for intelligent vehicle mass production.

　　Apollo Cloud 2.0 Framework

　　Apollo Cloud 2.0 is equipped with 12 core technologies, such as a textual model, a simulation engine, automatic annotation, model training, data playback, unlimited mileage, difficult case mining, and workflow engine. It is equipped with massive data resources, enabling application scenarios such as data compliance at the mass production stage of automobile enterprises, massive data mining and purification, urban level simulation, algorithm research and development, and vehicle operation supervision. Through integrated services, autonomous driving development has become smarter, more efficient, and simpler. Powered vehicle enterprises have achieved "from having to optimizing" autonomous driving, achieving data closed-loop, compliance closed-loop, and scenario closed-loop, and seizing the lead in the intelligent driving service market.

　　"Three Core Competencies" Assist in Mass Production Research and Development of Smart Cars to Reduce Costs and Improve Efficiency

　　Apollo Cloud 2.0 has three core service capabilities: precision compliance, multimodal large models, and city level simulation services, which are key to solving the long tail problem of intelligent driving and helping promote the mass production process.

　　"Accurate compliance" is an innovative strategy pioneered by Apollo Cloud 2.0, which not only avoids the waste of resources caused by "excessive compliance", but also eliminates the drawbacks of "on-demand compliance", which can lead to headaches and foot pain. It can better meet the development needs of intelligent driving business. "Accurate compliance" effectively ensures the balance between cost and compliance, comprehensively ensuring compliance control of autonomous driving data closed-loop, and achieving full control of original data without vehicle delivery, mapping data without cloud delivery, mapping results without correlation, and qualification map providers.

　　The construction of Baidu's "precision compliance" service capability stems from its rich experience in surveying and mapping geographic information data processing, followed by a deep understanding of the purpose and bottom line of compliance requirements, and finally, its insight into the business and needs of automotive enterprises. Baidu has the only compliance team in the industry that understands both compliance and safety, and is proficient in the autonomous driving business: a surveying and mapping security committee composed of a surveying and mapping security team, an information security team, and an autonomous driving compliance business team, which can provide full life cycle security services and multi department "consultation" services to escort the development of intelligent driving for automotive enterprises.

　　Quickly and accurately mining valuable information from massive data is the key to affecting the iteration speed of intelligent driving. The traditional manual labeling process is complex and inefficient, which has become a bottleneck in the industry. The "multimodal big model" of Apollo Cloud 2.0 upgrades data services from "process based" to "retrieval based". Using the capabilities of the Wenxin big model, it builds an autonomous version of the "data intelligence search engine", realizing the "needle in a haystack" of data mining, and can quickly and accurately mine valuable information from massive data, effectively improving the iteration speed of intelligent driving. In the past, a special scenario data mining task was conducted through manual annotation, which took a week to develop. Now, vehicle companies can access it with one click.

　　It is well known that test vehicles can run in cities, but mass production vehicles need to run out of cities. For intelligent driving research and development vehicle companies lacking data accumulation, how to quickly conduct research, development, testing, and operation of autonomous driving in a low-cost manner, and the application of high-precision, multi-scenario city-level simulation is the key.

　　Apollo Cloud 2.0 City Level Simulation Cloud Service

　　Apollo Cloud 2.0 provides automotive companies with over 100 cities of real road networks, tens of thousands of kilometers of scene data, and supports large-scale simulation tests that travel tens of thousands of kilometers per day. Over the years, Baidu has accumulated a large amount of automatic driving scene data, which on the one hand comes from the twin cities generated by Baidu Map's large-scale road network; On the other hand, it comes from the massive scene of Apollo's tens of millions of kilometers of road test mileage accumulation. With the help of Apollo Cloud 2.0, car companies can easily verify their autonomous driving capabilities in different cities and scenarios in the cloud, solving the problem of "difficulty in getting smart cars out of the city" caused by differences in geographical environment and road conditions. By releasing Apollo's billions of valuable data, car companies can quickly accumulate testing mileage, significantly reduce research and development costs, and improve research and development efficiency by more than 10 times.

　　Baidu has been committed to promoting the intelligent development of the automotive industry. In Gao Guorong's view, smart cars are about to experience a real explosion, with increasing demand for mass production. The "butler style" cloud service provided by Apollo Cloud 2.0 can effectively solve the many challenges faced by smart car mass production. Baidu is also looking forward to working with industry partners to overcome difficulties in autonomous driving technology, seize growth opportunities, and jointly embark on a new era of intelligent driving mass production.