Recently, Baidu Create 2021 (Baidu AI Developer Conference) Autonomous Driving Sub-forum "Creating Auto Robot Connection Technology and Scenarios Together" was held in the "Xiyang" APP. At the forum, Baidu Apollo Autonomous Driving Open Platform Apollo7.0 version was officially released. It also showed the outside world the layout of the Apollo X plan in the ecology of multiple car robots. Baidu Apollo is accelerating the development and commercialization of China's autonomous driving industry through the use of diverse automotive robot connection technologies and scenarios, and using technology to empower ecological partners.
The self-driving open platform is an important support for Baidu's multi-car robots. The Apollo7.0 released this time has achieved a milestone complete evolution from code to tools, and from open source platforms to tool-based platforms. Based on the four open source platforms of cloud services, open source software, hardware development, and vehicle certification, Apollo 7.0 provides a series of upgrades including one-stop practice cloud platform Apollo Studio, industry-leading simulation services, and efficient new models. Not only code Fully open and capable of providing a full-stack tool chain for autonomous driving, which is easier to use, more advanced, and more efficient to help developers use platform capabilities.
At the cloud service platform level, Apollo 7.0 officially upgrades the "data pipeline" service popular among developers in version 6.0 to Apollo Studio, covering the entire process of cloud tool chain for developers from on-board to on-board practice, providing developers with one Station-based practice platform experience. At the simulation platform level, Apollo 7.0 launched the industry's first PnC reinforcement learning model training and simulation evaluation platform. It has multiple advantages such as real data, powerful functions, comprehensive evaluation standards, and scalable architecture. It is expected to provide a unified verification standard for reinforcement learning research. At the open source software platform level, Apollo7.0 upgraded the perception and prediction algorithm modules, and introduced three deep learning-based models, MaskPillars, SMOKE, and Inter-TNT, to effectively reduce problems such as missed detection and jitter.
At present, Baidu’s Apollo autonomous driving open platform has completed 11 iterations. It has more than 80,000 developers and more than 210 partners in 135 countries around the world. The total amount of open source code is 700,000 lines. It has grown into the world's most active autonomous driving open. platform.
In addition, Baidu Apollo also shared with developers the rich resources and latest developments of the Apollo EDU Autonomous Driving Education Talent Program, and announced the goal of empowering tens of millions of colleges and enterprises and cultivating millions of autonomous driving engineering talents in the next three years. , Comprehensive coverage of application-oriented and research-oriented talents, enabling autonomous driving to truly empower all walks of life and promote the continued prosperity and development of the autonomous driving industry.