Article directory
Today, with the rapid development of autonomous driving technology, becoming a developer in this field is a challenge, an adventure, and a spiritual journey. As one of the pioneers in this field, Apollo Open Platform 9.0 was released on December 19. At the same time, Apollo Open Platform 9.0 provides developers with a systematic and comprehensive introductory development guide, making it very easy for developers who are new to this field to get started. The following is Let me share with you my growth and experience in building the platform.
Platform architecture:
Because the project needs to build an autonomous driving system platform, we conducted research on community activity, framework tools, documentation tutorials, platform functions, etc. and finally selected the Apollo open platform as our autonomous driving system development platform. The first task when building is to understand the platform architecture. It consists of four layers: hardware equipment platform, software core platform, software application platform and tool service platform.
From the bottom up, the hardware equipment platform mainly solves the problems of wire-controlled vehicles and sensors and other hardware equipment in the process of building the autonomous driving system. The software core platform provides the autonomous driving vehicle software system framework and technology stack. It includes the underlying operating system, the real-time communication framework in the middle layer, and the upper autonomous driving application layer, such as perception, prediction, planning, control, positioning, etc. The software application platform is oriented to the engineering of different application scenarios and the capability expansion of autonomous driving application modules. Through the application platform layer, developers can more conveniently tailor, combine and expand based on the capabilities of each module of the platform. The tool service platform provides R&D infrastructure in the autonomous driving R&D process and improves autonomous driving R&D efficiency.
The biggest pain point in the development of autonomous driving systems is the cost of actual testing, and the amount of data is also very large. The Apollo open platform cloud service platform solves the problem of data utilization efficiency through the cloud, and reduces the cost of actual vehicle testing by combining with simulation, which can greatly Improve the efficiency of autonomous driving research and development based on the Apollo open platform. This is also the reason why we chose the Apollo open platform as our autonomous driving system platform.
Basic environment:
The Apollo open platform requires the installation of necessary basic software before it can be developed and run. Since the Linux system is divided into multiple branches such as Redhat, Centos, Fedora, etc., when choosing a system, it is best to choose the officially recommended Ubuntu 18.04 as the basic environment, because it has been officially verified Also to avoid pitfalls. The Apollo open platform is very convenient to start and manage based on the Docker container. There is no need to learn how to install Docker, just run the official script directly.
The environment management tool of Apollo open platform 9.0 can help us manage and start platform environment containers. The installation is also simple and just follow the community documentation.
start using:
The officially provided Dreamview+ function is very powerful and can visually display the output information of the current autonomous vehicle module. For example: planned path, vehicle positioning, frame information, etc.
Provide users with a human-computer interaction interface to monitor vehicle hardware status, switch modules on and off, start autonomous vehicles, etc. Provide debugging tools. For example: PnC monitor can efficiently track problems with module output. However, most of the functions of Dreamview+ are mainly for use in actual vehicle debugging.
Developers who develop perception modules are provided with data operation process options, visual data display panels and debugging information panels related to perception development and debugging. The PnC development and debugging mode is suitable for developers who develop planning and control modules. It provides data operation process options, visual data display panels and debugging information panels related to PnC development and debugging.
Dreamview+ helps us run and monitor the effects of planning and control algorithms in scene simulation. It can set the opening and closing of planning, control, and routing algorithms. It can synchronize the scene from the cloud and run it. It can add different driving trajectories to the scene and can be run during the operation. View the module delay, console logs, data records of the PnC algorithm through the monitoring module, and message information in each channel. You only need to download the required scene set from the Apollo Studio cloud and select the scene to be run in the local scene list, then select the vehicle in the autonomous driving system resources, open the module, draw the trajectory, and run the simulation to complete. This function can help us simulate various complex road conditions. Then check the module delay, console and other information and debug the code according to your own needs.
Experience:
Apollo开放平台9.0提供了全面的开发工具和文档。通过详细的文档，我们可以轻松了解整个平台的架构、功能模块以及使用方法。Apollo开放平台的开发工具涵盖了传感器模拟、地图数据管理、车辆控制等方面，让开发者可以一站式完成整个自动驾驶系统的构建。文档的详细程度和示例代码的完备性为开发者提供了强有力的支持，大大降低了学习和使用的门槛。
In the community, you can also communicate with peers from all over the world, share experiences, and solve problems. This open communication atmosphere provides convenience for solving practical problems and also promotes the continuous progress of autonomous driving technology. In my experience, community support allows me to resolve difficulties more quickly and receive feedback and suggestions.
The modular design of Apollo Open Platform 9.0 makes system construction more flexible. We can select appropriate modules for integration according to project requirements, avoiding development from scratch and improving development efficiency. The platform supports the input and output of multiple sensor data, making the system suitable for different scenarios and models. This flexibility allows us to better adapt to diverse application scenarios and improves the scalability of the system.
Overall, I deeply appreciate the powerful functions of the platform, the friendly development environment and the rich community support. As autonomous driving technology continues to develop, I believe the Apollo open platform will continue to play an important role in promoting innovation and progress in the field of autonomous driving.