Apollo Open Platform 8.0 released: multi-dimensional upgrade "born for developers"

Apollo Open Platform 8.0 Released: Multi-dimensional Upgrade "Born for Developers"
Apollo Open Platform Welcomes Version 8.0, Baidu Autopilot Open Platform Enters the Era of Ease of
Use More than 10,000, more than 700 colleges and universities
Apollo Studio learning practice community is online, and it only takes 5 minutes for a novice to run through a Case

On December 28, Baidu held the Apollo Open Platform 8.0 online conference. At the meeting, Baidu officially launched a new upgraded version of the Apollo autonomous driving open platform for all developers - Apollo Open Platform 8.0, which further consolidated the ease of use of the platform and made it easier for developers to operate. At the same time, Baidu Apollo also shared the latest progress in autonomous driving education and ecological partners with the outside world.

The official release of Apollo Open Platform 8.0 brings a new multi-dimensional upgrade

The Baidu Apollo open platform has always maintained innovation and high-frequency self-iteration since its launch. From the beginning, it focused on the construction of basic capabilities, to the development of scenario capabilities, and then upgraded to the pursuit of excellence in the ease of use of platform engineering. Zhang Liang, head of Baidu's autonomous driving platform ecology, said, "Apollo Open Platform 8.0 has been comprehensively upgraded from the two important levels of 'new architecture' and 'new capabilities', and has been improved based on the actual needs of developers to help developers more. Better and faster to get familiar with and use Baidu Apollo open platform."

First of all, at the platform architecture level, the Apollo Open Platform 8.0 has been upgraded from a technology-oriented layered architecture to a layered architecture that combines technology and ecology. In the "new architecture" system, the bottom layer is the hardware device, which can help Apollo's ecological partners to provide developers with a rich selection of autonomous driving devices through open standard protocols and certification. On top of this is the software core, which can provide the ability to build the smallest software subset based on the Apollo autopilot system. This upgrade opens up the entire process of perception development and provides PnC simulation debugging capabilities based on the local HMI Dreamview, thereby improving perception and PnC R&D efficiency. The next level up is the software application, which can provide developers with brand-new autopilot module expansion capabilities and expansion methods, and open it in the form of scenarios. At the top of the architecture is the cloud service, which has added model training service capabilities, and upgraded the training platform and simulation capabilities. Through this new architecture, Apollo Open Platform 8.0 can provide developers with easy-to-use software core and cloud service tools, as well as easy-to-expand hardware devices and software scene applications.

Secondly, in terms of platform capability building, the Apollo Open Platform 8.0 has been upgraded mainly in three aspects. First, introduce a software package management mechanism on the engineering framework to organize and manage Apollo modules in the form of software packages, so that developers can download on demand, reduce the amount of data, and shorten the time required for installation and deployment from "days" to Within 30 minutes, the threshold for installation and engineering expansion is greatly reduced. Second, Apollo Open Platform 8.0 has upgraded the perception framework and development process on the basis of the original. On the one hand, it has introduced three new models based on deep learning, which further enriches the types of perception models and effectively improves the accuracy; In the whole-process development process of perception, from model training, deployment to verification, developers can not only choose the model that comes with the platform, but also expand more models based on their own needs. Third, Apollo Open Platform 8.0 adopts a new PnC tool chain, which can provide local debugging function of PnC simulation test and convenient cloud simulation scene management function. Developers can download the scenes and dynamic models of Apollo Open Platform through plug-ins To the local, the overall debugging efficiency can be more than doubled.

In addition, another initiative of Apollo Open Platform 8.0 is the launch of a new one-stop learning practice community for autonomous driving - Apollo Studio. For autonomous driving, a field with high technical threshold, difficult learning, and high practical operation cost, Apollo Studio can help developers who are waiting to see, want to enter, or want to seek breakthroughs to solve some problems. This community can provide developers with opportunities to learn courses and practice on the cloud. It only takes 5 minutes for a novice to run through a sample case, which greatly reduces the difficulty of learning. At the same time, it can also provide developers with a series of automatic driving PnC, perception Event opportunities allow developers to compete and challenge themselves with their colleagues around the world; and, as a community that gathers experts in the autonomous driving industry, it can become a place for developers to conduct in-depth technical exchanges with industry leaders, and An excellent window for peer sharing and interaction.

Apollo EDU's phased results are outstanding and will join hands with BIT to start a new cooperation

A very important part of Baidu's Apollo ecosystem is education and talent training in the field of autonomous driving. In the past few years, Apollo has continued to join hands with educational ecological partners to cultivate interdisciplinary talents with cross-integration capabilities and innovative practice capabilities to "create blood" for the industry. In December last year, Baidu released the Apollo EDU talent training plan, setting the goal of training one million autonomous driving talents in three years, covering thousands of colleges and universities. At present, through various practical actions such as school-enterprise cooperation, teacher training, and holding competitions, Apollo has fulfilled the previously set goals in stages. Apollo EDU has covered more than 335,000 technical talents related to autonomous driving, covering colleges and universities The number reaches more than 700.

In terms of industry-university-research cooperation with authoritative organizations, Apollo also joined hands with the China Computer Federation Intelligent Vehicle Branch (CCF IV) last year to build the "Autonomous Driving National New Generation Artificial Intelligence Open Innovation Platform" to conduct technical seminars for developers, Science Education Lecture Hall, etc. At this press conference, Apollo further announced that it will further cooperate with Beijing Institute of Technology to jointly build a joint innovation center for intelligent connected vehicles. In the future, it will work together to cultivate high-quality compound talents for intelligent connected vehicles.

In fact, in addition to the above-mentioned intensive measures, Apollo EDU has also explored a set of effective solutions for the cultivation of talents in the autonomous driving industry, so as to ensure that different types of talents are "taught in accordance with their aptitude". For college talents, Apollo adheres to the concept of integrating production, learning, research, and application, and provides colleges and universities with a virtual teaching and research room platform and a full-link solution for the DKIT on-board practice platform. For industry-oriented talents, it focuses on providing practical and practical course systems to cultivate application-oriented talents such as autonomous driving integration, testing, and operation and maintenance, as well as research-oriented talents such as algorithms and engineering technologies. Zhang Liang said, "In the future, Apollo EDU will also cooperate with more universities and institutions in education, so that more developers who are interested in autonomous driving will have the opportunity to enter the industry, and at the same time, it will also cultivate a steady stream of new forces for the industry."

Apollo X proposes a targeted co-promotion plan to help enterprises implement large-scale business

The Apollo Open Platform also proposed a "Map for Building Autonomous Driving Capabilities for Enterprises" specially designed for ecological partners at this press conference. Based on the stage characteristics of the enterprise's own scenario application, different co-promotion schemes are provided to help the enterprise realize the large-scale commercial implementation of autonomous driving better and more efficiently.

For enterprises in the early stage of prototype construction, the solution proposed by Apollo X mainly starts with the tools and capabilities required for construction, and provides enterprises with domain control including Apollo core system, secondary development interface, and functional safety software capabilities. Kits including sensors and recommended sensors, as well as end-to-end construction services and technical training consultation.

For companies that are in the middle and late stages, have small-scale mass production capabilities, and start scene operations, the solutions provided by Apollo X mainly focus on software capabilities that match the application scenarios, as well as simulations supported by large-scale computing power, Model training, data management and other research and development tool platforms and business monitoring cloud tools, etc., mainly help enterprises optimize operating costs, risks and efficiency from the perspective of large-scale and mature scenario operations.

The Apollo open platform has always adhered to the core concept of "open capabilities, shared resources, accelerated innovation, and continuous win-win", constantly bringing innovation and upgrades. The launch of the Apollo Open Platform 8.0 once again makes the Apollo Open Platform a big step forward in terms of engineering ease of use, lowering the threshold of operation difficulty and operating cost, so that more developers can easily and conveniently get started with the Apollo Open Platform, Into the field of autonomous driving technology.

At present, the amount of open source code of the Apollo open platform has exceeded 750,000 lines, and has brought together more than 100,000 developers from 165 countries around the world, and has more than 220 ecological partners around the world. In the future, Apollo also looks forward to working with more developers and partners to continuously expand the boundaries of capabilities, explore more possibilities of autonomous driving technology and the industry, and work together to "new and new" and grow together.