On February 24, the world's first vehicle-road collaborative autonomous driving dataset BASED ON REAL SCENEs WASAIR-V2X was officially released, providing domestic users with download and use. The dataset was jointly released by the Intelligent Industry Research Institute (AIR) of Tsinghua University, beijing high-level autonomous driving demonstration zone, Beijing Auto Network Technology Development Co., Ltd., Baidu Apollo, and Beijing Zhiyuan Artificial Intelligence Research Institute. The dataset released this time for the first time realizes the innovation of 2D and 3D marking methods from the joint perspective of the car end and the road end in the same time and space, and as the first open source vehicle-road collaborative dataset in the industry and academia, it will effectively serve scientific research, industry and government agencies, effectively coordinate with all parties to carry out academic research and industrial landing of vehicle-road collaboration, and promote the coordinated development of mainland vehicles and roads.
(Press Conference)
The dataset released this time comes from 10 kilometers of real urban roads, 10 kilometers of highways, and 28 intersections in the Beijing High-level Autonomous Driving Demonstration Zone, including 71254 frames of image data and 71254 frames of point cloud data from various types of sensors such as car end, road end camera and car end, road end lidar, etc., covering rich scenes such as sunny, rainy, foggy, day and night, urban roads and highways. Compared with the dataset that contains only the single-vehicle end or the single-way end, the dataset provides multimodal data of the joint perspective of the lower end of the car and the end of the road in the same time and space, and provides the results of fusion labeling under the joint perspective of different sensors, which is used to better serve the research and evaluation of the vehicle-road collaborative algorithm. In addition, the dataset effectively reduces the cost of dataset construction through innovations such as semi-automatic self-learning vehicle-road collaborative 3D fusion annotation method.
(Kong Lei, Deputy Director of the Administrative Committee of Beijing Economic and Technological Development Zone and Director of the Beijing High-level Autonomous Driving Demonstration Zone Office)
High-quality data is the key to vehicle-road collaborative automatic driving technology, which can continuously optimize the algorithm of vehicle-road collaboration, help the iterative upgrading of automatic driving systems, and make a good safety guarantee for the large-scale promotion of high-level automatic driving. At present, Beijing is seizing new opportunities for the coordinated development of vehicles and roads, with a forward-looking strategic vision, building the world's first networked cloud-controlled high-level autonomous driving demonstration area and the first intelligent networked vehicle policy pilot zone in China with high standards, creating a test technology sandbox for autonomous driving technology research and development research institutions and scientific and technological enterprises, and providing policy, regulation and technical specification support. The release of this dataset brings massive, multi-modal, multi-perspective real scene data to the industry and academia, and publishes the data after processing such as labeling, desensitization and security encryption, so as to promote the academic and industrial circles to jointly promote data-driven vehicle-road collaborative autonomous driving. Kong Lei, deputy director of the Management Committee of Beijing Economic and Technological Development Zone and director of the Beijing High-level Autonomous Driving Demonstration Zone Office, said, "In the future, we will use data transformation to formulate a number of vehicle-road collaborative data standards in a timely manner, promote the unification of industry data elements, interfaces, formats and other standards, and provide reference and guidance for the construction of demonstration areas and industry development; through data opening, provide basic data for universities and scientific research institutions, provide support for enterprise product research and development testing, and effectively accelerate the collaboration between production, education, research and application. At the same time, we will continue to explore data openness and shared service models to promote the development of the digital economy. ”
(Professor Guo Qiang of Tsinghua University, Researcher of Intelligent Industry Research Institute Nie Zaiqing on the spot)
In February 2020, the National Development and Reform Commission and relevant ministries and commissions issued the "Intelligent Vehicle Innovation Development Strategy", which regarded "promoting the planning and construction of intelligent road infrastructure" as an important national strategic task, and clarified the autonomous driving route with Chinese characteristics of "bicycle intelligence + vehicle-road collaboration", and vehicle-road collaboration has become the focus of research from all walks of life. Based on this, Zhang Yaqin, dean of the Intelligent Industry Research Institute (AIR) of Tsinghua University, gave his own judgment, he believes that the construction of high-level intelligent networked roads is a major trend of intelligent driving and intelligent transportation, which not only ensures the possibility of mass production of intelligent vehicles, but also ensures the feasibility of high-level automatic driving based on scenario-driven landing. In the field of intelligent transportation, the Intelligent Industry Research Institute (AIR) of Tsinghua University has carried out school-enterprise cooperation with a number of enterprises, including vehicle-road collaborative autonomous driving, vertical industries, trolley logistics, Robotaxi and other application scenarios. In the face of the lack of relevant data sets, can not meet the real problems of data-driven vehicle-road collaboration from all walks of life, Tsinghua University Intelligent Industry Research Institute (AIR) initiated the construction of data sets, and opened the data sets to universities, research institutes, and industrial users, which can better support scientific research institutions to conduct scientific research; at the same time, the real data collection, diverse scene coverage, and massive data samples in the demonstration area can also solve the practical needs of enterprises in product development.
Shang Guobin, vice president of Baidu and general manager of the intelligent transportation business unit, said in his speech: "Data is the driving force for breakthroughs in the fields of autonomous driving, vehicle-road collaboration and intelligent transportation. The effective accumulation of data can not only qualitatively improve the effect under the conditions of the same level of algorithm and computing power, but also automatically identify and solve the problem, forming a positive cycle of artificial intelligence 'more knowledgeable, more and more intelligent'. "Baidu, as the technical supporter of the construction of the vehicle-road collaborative data set, will continue to work on building a sound vehicle-road collaborative ecosystem, and will gradually support the release of relevant tools and codes that are easy for users to get started and benchmark reproduction, and carry out data set-related use training to support the academic community to carry out research on vehicle-road collaboration."
The release of the world's first autonomous vehicle-to-road collaborative dataset, THE AIR-V2X, is of great significance for promoting the research and development of high-level autonomous driving technology in mainland China. At present, the dataset has been incorporated into the Zhiyuan platform, and the follow-up will rely on the Zhiyuan community and other Zhiyuan academic ecological networks to accelerate the opening, promotion and application of the dataset for all parties involved in industry, education and research.
Under the guidance of the strategy of new infrastructure and transportation power, the autonomous driving route with Chinese characteristics of "bicycle intelligence + vehicle-road coordination" has gradually gained international recognition with its leading nature. In the future, under the effective collaboration of all walks of life based on the vehicle-road collaborative automatic driving data set, it will continue to promote the application of vehicle-road collaborative automatic driving and help the rapid development of the mainland's autonomous driving and intelligent transportation industry.