On December 19th, the new and upgraded version of the Apollo open platform was officially launched - Apollo open platform 90，To meet the needs of all developers and ecosystem partners, with stronger algorithm capabilities, more flexible and easy-to-use tool frameworks, and more easily expandable general scenario capabilitiesWe will continue to build a leading edge in the development of autonomous driving. At the meeting, Apollo also shared the latest progress in autonomous driving education and ecological partners.
Since its launch, the Apollo open platform has maintained high-frequency iterations, from the continuous optimization of the engineering framework, to the flexibility and ease of use of developers, to the ability to implement general scenarios, and has always broken through the boundaries of capabilities. Zhang Liang, general manager of the ecological department of the autonomous driving platform, summarized the main highlights of the upgrade: "Apollo Open Platform 90 has achieved a comprehensive upgrade in engineering, algorithms, and tools, and the general layer can enable the large-scale implementation of a variety of application scenariosThe overall operation is more flexible and easy to use, and the use scenarios are universal and easy to expand. While greatly improving development efficiency, it can help more developers quickly build their own autonomous driving systems. ”
In terms of engineering framework, in order to enable developers to assemble autonomous driving applications more flexibly and redevelop them more conveniently, the Apollo Open Platform 90 comprehensive upgrade of package management,By splitting modules into smaller packages according to functional granularity, developers can more easily choose and use them according to their needs. At the same time, it also provides a wealth of functional components and plug-ins, and improves and optimizes the function extension. Based on this, after the unified scheduling interface, developers can complete the scene demo construction within 1 day at the earliest, simplify the parameter adjustment method to increase the parameter adjustment efficiency by 1 times, and add a plug-in mechanism to reduce the learning cost by 90% and the amount of the Xi by 50%, which greatly improves the secondary development capability of Apollo. Apollo Open Platform 90 also adapts to the ARM architecture for the first time.
In terms of algorithms, the Apollo open platform 90 optimizes the perception algorithm on the basis of the previous one, LiDAR detection adopts the relatively new CenterPoint model, and visually adopts the YOLO X + YOLO 3D model, and the two models are trained with millions of data, and the recall rate and accuracy are greatly improved. In addition, it provides incremental training to support independent model training, which can significantly improve the detection ability of special targets and special scenes on the premise of maintaining the original detection capability of the model, so as to easily improve the detection effect of customized scenarios at a lower cost. In addition, it fully supports 4D millimeter-wave radar, which greatly enhances obstacle detection and safety in extreme weather scenarios.
In terms of tools, in addition to the new functions on the basis of the original tools, new tools including high-definition mapping, sensor calibration and integration have also been added. Apollo Open Platform 90The newly upgraded DreamView+ has been comprehensively improved in many aspects such as multi-scene use, free layout, and data resources, and the debugging process is more concise, the window layout is more flexible, and the resource access is more convenient.
In addition, the Apollo Open Platform 90 also reconstructs the document platform to make the operation more convenient, the reading smoother, and the content more substantial, effectively reducing the learning Xi and use costs for developers.
At the press conference, Dr. Dong Zhaozhi, Executive Vice President of Kaiwo Group and President of the Research Institute, Qiang Xiaowen, Secretary of the Party Committee, Deputy Director of the Business Research Headquarters and Dean of the Commodity R&D Institute of Dongfeng Motor Co., Ltd., Dr. Liu Mingchun, Vice President of the Prospective Technology Research Institute of Jinlong United Suzhou and Dr. Liu Junchuan, Vice President of TZTEK Technology Co., Ltd., also attended the scene and expressed their personal experience on the development and industry application of autonomous driving.
There are a large number of subdivided scenarios in the field of commercial vehicles, but there is a certain gap between the number of uses compared with passenger cars, and how to quickly deploy full-stack technology for autonomous driving under the premise of controlling costs and ensuring safety is still a difficult problem encountered by industry participants.
Apollo Open Platform 90 has powerful general-purpose capabilities, and is easier to use and expand. The number of adaptation links can be reduced by 40%, the number of readings reduced by 90%, and the amount of debugging can be reduced by 80%, which can achieve "out-of-the-box use" and complete the closed loop of autonomous vehicles within one week. At the same time, the Apollo open platform 90 coverage scenarios are more and more abundant, and the closed-loop operation of the scenario application business system can be completed in 1 month. On this basis, the sensor calibration and map creation cycle are shortened to hours, which greatly shortens the landing time and is more efficient.
At the same time, the hardware cost is lower, the selection is more abundant, the hardware selection supports 3LIDAR+4Camera, and the camera supports more than 4 manufacturers, from USB30 upgraded to GMSL, LiDAR has added 32-line, 64-line and other multi-brand and multi-model equipment, and more than 3 mainstream brand equipment has been added to the positioning equipment, which can help the large-scale application of different scenarios in an all-round way.
Apollo Open Platform 9The general capability of 0 can help partners such as traditional manufacturing car companies build their own autonomous driving systems for different scenarios and applications in the wave of intelligence, and solve problems such as difficult system construction, high development costs, less hardware support, and low tool efficiency. By building a new paradigm of all-round cooperation such as talent training, technical training, and business traction, partners will have the confidence and ability to face the intelligent transformation and upgrading of future products.
As a pioneer in the field of autonomous driving, Apollo attaches great importance to education and talent training in the field of autonomous driving. Through the Apollo Studio developer community, a one-stop learning Xi practice platform for courses, experiments, and competitions has been built. At present, the number of Xi students in the apollo studio developer community has exceeded 380,000, covering a total of 68% of science and engineering colleges and universities offering autonomous driving-related majors across the country, becoming the largest and most influential community in China for autonomous driving training.
In the "China Robotics and Artificial Intelligence Competition" hosted by the Chinese Society of Artificial Intelligence, apollo's competition has received extensive attention and recognition from teachers and students of colleges and universities and the ** society, and the two competitions have attracted more than 5,000 people from more than 1,500 teams from more than 300 colleges and universities, which is currently the largest registration competition in the national autonomous driving category.
Tan Qingji, Secretary General of the China Robotics and Artificial Intelligence Competition, pointed out: "The in-depth exchanges and cooperation between apollo and the Chinese Society of Artificial Intelligence is an important step in jointly promoting the process of scientific and technological talent training. Through such cooperation, we hope to cultivate more innovative and practical scientific and technological talents in the new era, and jointly promote the development of China's intelligent technology. ”
At the press conference, the Apollo EDU university plan was officially upgraded, and a three-dimensional school-enterprise cooperation talent training solution was built for the undergraduate level, and cooperation was carried out in teaching, teachers, innovation, training, cooperation support, etc. At the same time, at the vocational education level, a technical and skilled talent training program has been created around the job skill map of intelligent networked vehicles.
Wang Wenkai, Deputy Secretary of the Party Committee and Dean of the School of Traffic Engineering, Nanjing Polytechnic University, said: "Since 2021, the school and Apollo have jointly explored a characteristic development path of vocational undergraduate education from the aspects of curriculum construction, faculty construction, scientific and technological innovation competitions, and practical Xi employment, and have achieved positive results. In the future, the two sides will continue to further promote the cultivation of professional talents related to intelligent networked vehicles, and cultivate more high-level technical and skilled talents in the intelligent networked vehicle industry. ”
At present, the Apollo Open Platform brings together more than 160,000 developers from more than 170 countries and regions around the world0 to 90During the development process, 120,000 rows** were refactored and 200,000 rows** were added. The Apollo open platform 9The launch of 0 makes it more complete and rich in functions, more flexible and easy to use in operation, and easy to expand in scenarios. In the future, the Apollo open platform will focus on the individual needs of each developer and ecological partner, continuously expand the boundaries of capabilities, improve ease of use, work with developers and partners to create value, and run a new speed of autonomous driving applications.