On December 2, 2023, I was fortunate enough to participate in a technical exchange event held by Baidu Apollo. Apollo is an open source autonomous driving project developed by Baidu. This offline technical exchange gave me an understanding of autonomous driving. Deep understanding.
During the event, experts from the Apollo project introduced and demonstrated Apollo technology. They explained the technical architecture evolution, functions and application scenarios of the Apollo platform in detail and patiently, and through the demonstration of the cloud experiment platform, I felt the future of autonomous driving technology. I learned that Apollo Beta has comprehensively upgraded engineering, perception algorithms, developer tools, etc. on the basis of 8.0, and can realize full-process autonomous driving from vehicle perception, decision-making to execution.
Apollo Beta has comprehensively upgraded the project on the basis of 8.0. Apollo 8.0 introduced the concept of software package management for the first time, organized various Apollo modules and released them in the form of software packages, optimized the download and installation process and improved development efficiency. . Now, the new version of Apollo Beta launches package management 2.0, which splits modules into smaller software packages according to functional granularity. Developers can freely choose and combine according to their needs. Not only that, Package Management 2.0 has also optimized and improved functional expansion, greatly improving Apollo’s secondary development capabilities.
Beta has upgraded and optimized the perception algorithm based on Apollo 8.0. Lidar detection adopts the relatively new CenterPoint model and uses Baidu’s million-level data for training. Visually, it uses Yolo X and Yolo 3D models. The detection effect and generalization have been greatly improved, and incremental training is also provided to support independent model training.
In addition, Beta perception adds support for 4D millimeter waves, and obstacle detection and weather adaptability have been greatly enhanced.
In the new version Beta, a series of technical upgrades and functional expansions have been made to the Apollo developer tool Dreamview, bringing a new Dreamview+. The new version of Dreamview+ has been comprehensively improved in four aspects: multi-scenario use, free layout, data resources, and language support:
- Mode-based multiple scenarios - the process is more concise: specific development scenarios such as sensing and PnC are used as mode classifications to streamline the operating steps in various modes, optimize the use process, and improve development efficiency;
- Panel-based layout - visualization is more flexible: supports free configuration of the layout, content and size of each panel, allowing developers to create an operation interface that matches their own workflow;
- Integrated cloud resource center - more convenient access: centralized data resources can be easily managed and quickly accessed. At the same time, the interaction with Studio cloud resources can be further enhanced. Various resources can be downloaded for algorithm testing with one click, including maps, scenes, and vehicles. Configuration, data packages, etc. to further enrich the developer resource library.
- Newcomer guidance and switching between Chinese and English - making it easier to get started: Visual guidance is provided to help new developers quickly learn the operating procedures and steps, reducing learning and exploration costs; all functions support switching between Chinese and English, reducing the difficulty of understanding professional terms and responding to domestic and foreign responses. Developer demands.
After the experts’ speeches, the organizers arranged free discussion time, and I communicated and interacted with other participants. This gave me the opportunity to meet autonomous driving-related professionals from different units and share my insights and feelings with them. We communicated about the possible future development trends, challenges and effective solutions of autonomous driving technology. Through communicating with them, I expanded my horizons and gained some valuable advice and inspiration.
Participating in this Apollo technology exchange event gave me a more comprehensive understanding of autonomous driving technology. I deeply understand that the development of autonomous driving technology is driving changes in the transportation industry, and it will bring greater convenience and safety to people's travels. At the same time, I also realize that autonomous driving technology still faces many challenges and requires continuous innovation and breakthroughs.