Sharing of entry-level platform for autonomous driving technology: Baidu Apollo open platform 9.0 comprehensive upgrade

As programmers, we are looking for our suitable fields in every era. The rise of a new field can not only drive the development of related enterprises, but also drive employment, entrepreneurship, and development opportunities at the social level. Today, as we have entered the era of large AI models, the term "autonomous driving" stands out from many emerging technologies.

Baidu's Apollo open platform 9.0 provides such a platform for developers around the world. As a veteran company that has been involved in autonomous driving for more than 10 years, the release of version 9.0 of the Apollo open platform has made important breakthroughs in algorithm capabilities, engineering frameworks, scenario expansion, etc.! Below I will explain it to you step by step.

Comprehensive upgrade of the platform

Brand new architecture

The new architecture diagram is as follows. We can compare the architecture diagram of version 8.0. The first picture is the structure diagram of version 9.0, and the second picture is the structure diagram of version 8.0.

Version 9.0 has the following upgrades based on version 8.0:

Tool services

Model training
High-precision mapping tools
Sensor calibration
Integrated tools

Application software (scenario application)

Education, mining truck, environmental protection, logistics, inspection, connection, positioning, perception, prediction, planning, control, functional safety (new functions)

Software core

Perception, prediction, planning, control, HMI

hardware equipment

Reference hardware

Stronger algorithmic capabilities

Introduction of advanced algorithms: By introducing the latest algorithms such as advanced CenterPoint and Yolo X + Yolo 3D into the main model, developers can benefit from more powerful perception capabilities. This means that the system has better performance in target detection and multi-category scene processing, providing developers with more data and scene processing possibilities.
Deep training and large-scale data: The main model is deeply trained with millions of training data, and the accuracy of the model is significantly improved. This enables developers to achieve more reliable results in a variety of complex real-world scenarios, whether in urban environments or more challenging traffic conditions.
Lidar perception upgrade: The CenterPoint model is introduced, which provides higher accuracy and generalization capabilities compared to the previous CNNseg. This provides a more reliable basis for environmental awareness, making developers more confident in handling obstacles and environmental awareness.
Camera perception upgrade: Transitioning from the traditional Yolo model to the structure of Yolo X + Yolo 3D, introducing advanced three-dimensional perception capabilities. This change enhances the system's adaptability to multi-objective and multi-category scenarios, providing developers with more options for handling diverse scenarios.
Inference latency optimization: Especially for scenarios with high real-time requirements, version 9.0 focuses on reducing the overall inference latency. This enables models to run smoothly on Orin chips, providing developers with the ability to respond quickly in real-time applications.
System performance and efficiency improvements: Through a number of optimization efforts, version 9.0 has achieved impressive progress in system performance and efficiency. This allows developers to better meet various needs in actual production, thereby improving the availability of the entire system.

Through the upgrade of the main model, version 9.0 of the Apollo open platform not only fully realizes the capabilities of the open platform, but also successfully integrates its own scene applications with it, providing developers with more powerful tools and resources. In addition to the upgrade of the main model, version 9.0 also introduces innovative solutions for incremental training to better meet the needs of various application scenarios, while taking into account the specific needs of developers and partners, providing a better solution for perception systems. The result is more flexible paths and methods.

Let's take a typical incremental training example to illustrate. When developers use the Apollo open platform version 9.0, they use incremental training to apply it to specific scenarios. In this scenario, a rather special vehicle, a truck, was successfully detected through the model that senses obstacles on the left side. Unlike ordinary trucks, the vehicle has no rear body or compartment, only a flat pallet.

This incremental training method empowers developers to make targeted adjustments and optimizations to the system based on the needs of their specific scenarios. Developers can directionally improve the model to better adapt to special models or environments by introducing data from special scenarios. This flexibility and customizable incremental training enables the perception system to better adapt to various complex practical application scenarios and improves the system's adaptability and performance.

From a developer's perspective , this incremental training solution brings significant benefits. First of all, version 9.0 provides a personalized and accurate model training path, allowing developers to more flexibly respond to the challenges of specific scenarios. Secondly, through incremental training, developers can effectively improve the system and improve detection accuracy, thus enhancing the performance of the entire perception system. This enables developers to better meet the requirements of specific applications and provide more reliable autonomous driving solutions.

The incremental training mode is adopted, and about 2,000 frames of specific data are introduced. After training on this series of incremental data, the figure on the right shows the significantly improved detection results. In this figure, the detection accuracy and the definition of the entire obstacle boundary are very clearly characterized. This improvement has a significant helpful effect on enhancing the overall perception ability and providing effective assistance.
Version 9.0 of the Apollo Open Platform introduces a number of new sensor technologies, including the increasingly popular 4D millimeter wave. This technology is fully supported in version 9.0, covering all aspects from the underlying software driver to the functional application level of version 9.0. 4D millimeter waves have multiple advantages in applications, including cost advantages, comprehensive point source data, and excellent assistance to autonomous driving perception systems in specific environments, especially in severe weather conditions. The introduction of this advanced sensor injects new energy into the entire algorithm system, further improving the performance and robustness of the system.
For developers, these improvements and the introduction of new technologies bring significant benefits. First, the incremental training mode enables developers to effectively improve detection algorithms and improve accuracy and precision. By introducing specific data, developers can optimize the system in a targeted manner to provide more reliable perception capabilities.
The introduction of new sensor technologies provides developers with more tools and options. The superior performance of 4D millimeter waves allows developers to more comprehensively consider the perception requirements of autonomous driving systems, while also becoming more competitive in terms of cost-effectiveness. This provides developers with more flexible solutions to meet the needs of different scenarios and applications.

Version 9.0 of the Apollo open platform provides developers with more powerful tools and resources through technological innovation and system optimization to promote the continuous development of autonomous driving technology. This enables developers to build, optimize and customize autonomous driving systems more efficiently, contributing significantly to the progress of the entire industry.

Summary of algorithm upgrade in version 9.0

The main model is more advanced and more effective, and its accuracy rate significantly exceeds that of the old model.
Incremental training solves the problem of scene adaptability, a small amount of data can greatly improve scene effects, and open source high-quality code optimization helps model training to be completed independently.
It fully supports 4D millimeter wave radar, enriches cutting-edge sensor selection, improves detection results, and enhances safety in extreme weather scenes.

Easier-to-use engineering framework

In version 9.0 of the Apollo open platform, it not only continues the excellent package management capabilities of version 8.0, but also performs more refined processing of software modules. This sophisticated modular design provides developers with a higher degree of flexibility, allowing them to customize the software architecture more precisely, from the module level to application capabilities, specific functions, and specific function module (feature) levels, and further Improved the efficiency of software construction for autonomous driving systems.

In terms of application of the engineering framework, version 9.0 introduces three key new changes. These changes are designed to meet the needs of developers at different levels of development capabilities and take into account the diverse needs of learners and final engineering users.

The new version strengthens the modular design and provides developers with more flexible opportunities to customize and integrate functional modules. This makes it easier for developers to respond to project needs and meet their individual development needs regardless of their technical level. The splitting of modules and the ability to in-depth customization make the system easier to expand and maintain, providing developers with greater room for creativity.
Introduces more advanced application capabilities and provides more powerful functions and features. This allows developers to respond more flexibly to the complexity of autonomous driving systems, while also meeting the pursuit of more advanced functions by advanced developers. The enhanced functions and features of the new version provide developers with more tools to better solve practical problems while making the system more intelligent and efficient.
The new version focuses on the balance of learning and engineering applications, providing a smooth learning curve for both beginners and experienced engineers. Beginners can gradually master all aspects of the system, while experienced engineers can enjoy more customization and optimization options to meet the needs of different users in academic research and practical applications. This balance provides developers with a better user experience while promoting the advancement of autonomous driving technology.

The technical upgrade of Apollo open platform version 9.0 provides developers with many significant benefits, especially for developers with in-depth development needs:

Comprehensive and flexible interface encapsulation: For developers building autonomous driving application demonstrations or test vehicles, version 9.0 provides a complete interface encapsulation mode, allowing developers to easily call all software functions. This comprehensive and flexible interface design simplifies the development process and provides developers with excellent basic capabilities, allowing developers to focus more on the specific implementation of the application.
Systematic parameter adjustment method: For developers with deeper development needs, version 9.0 provides a systematic parameter adjustment method. Through structured processing of global and local parameters, developers can finely manage and control the performance and results of the entire autonomous driving software by adjusting parameters. This allows developers to better adapt to the needs of different scenarios and applications and achieve differentiated configurations.
Core module of plug-in management: In response to deeper development needs, version 9.0 has plug-in management of key functions in the core module. This allows developers to more conveniently and flexibly combine various plug-in capabilities to build complex autonomous driving applications for specific scenarios. This flexibility and composability provides developers with greater autonomy, allowing them to conduct in-depth development more efficiently.
Conveniences brought by engineering framework upgrade:

Fine-grained package management: Splitting software packages into finer granularity allows developers to select and assemble as needed anytime and anywhere, further improving the ease of use of the system.
Reduce secondary development costs: Quickly complete the application scenario demo construction, and reduce the time and cost of secondary development through efficient parameter configuration and adjustment.
Template interface realizes function expansion: Function expansion is realized through template interface, code learning cost is reduced by 90%, and the code volume is reduced by 50%. This makes it easier for developers to understand and extend existing functions, improving development efficiency.

These technical improvements make the Apollo open platform version 9.0 a more flexible, efficient, and powerful autonomous driving software construction platform. It not only meets the needs of developers at different levels, but also promotes the further development of autonomous driving technology.

Development tool DreamView+ newly upgraded

Version 9.0 also officially brings DreamView+, a new supporting tool for autonomous driving development. The new DreamView+ solves many problems and deficiencies left in many old development tools. The first is the overall introduction of the new DreamView+

The large picture mentioned in the above description shows the interface innovation of the new version of DreamView+, allowing developers to more efficiently complete the work that originally required switching and jumping between multiple pages. This updated design focuses on centralizing development tasks on a simple and efficient interface to provide a smoother development experience.

Task centralization and simplified operations: By concentrating key statistics and completed work on one interface, the frequency of switching and jumping between different pages is reduced. This design enables developers to obtain the required information more quickly, thus improving work efficiency.
Multi-role support and customization: By providing perception mode and PnC mode for different developer roles, the interface design takes into account the diverse needs of developers. This kind of personalization allows developers to focus more on the data and functions required for their specific roles, improving the adaptability and flexibility of the workflow.
Deeper system status analysis: The new PnC mode provides richer visual data, helping developers analyze system performance more comprehensively and in-depth, and discover and solve potential technical problems. This deep system state analysis helps optimize code and applications, improving development quality.
Improving development efficiency and experience: Overall, these interface improvements and functional optimizations enable developers to carry out development work more conveniently, reduce unnecessary steps, and improve development efficiency and the overall user experience.

The upgrade of the new version of DreamView+ not only focuses on the user interface, performance and appearance, but also deeply introduces highly flexible and convenient customization functions, which has a significant impact on developers when writing code and developing applications.

The broad coverage of these custom features allows developers to personalize it according to their unique workflow and development needs, thereby increasing work efficiency and flexibility.

Drag and drop function support: The new version of DreamView+ supports the drag and drop function of pages, structures and windows. This allows developers to re-adjust the page layout, component structure and window position through simple drag and drop operations. This flexibility allows developers to easily customize the interface to fit their personal workflow rather than being tied to a fixed layout.
Customized layout of page functions: In addition to the drag-and-drop function, the new version of DreamView+ also allows developers to customize the layout of the entire page functions. This means that developers can freely adjust the position and arrangement of features on the page according to their personal preferences and workflow. This high degree of flexibility allows developers to focus on core tasks and utilize workspace more efficiently.
Improve work efficiency and concentration: The introduction of this custom feature greatly improves work efficiency. Developers can configure the development environment according to personal preferences and work needs to make it more in line with individual requirements. This customized working environment allows developers to focus more on core development tasks without having to worry about the limitations of the interface structure, thus improving overall work focus and efficiency.
Meet personalized needs: Most importantly, the new version of DreamView+ provides a high degree of personalized customization options to meet the unique needs of each developer. Both the interface layout and function arrangement can be flexibly adjusted according to the developer's work preferences, providing developers with a development environment that is more in line with their personal habits.

The introduction of these custom functions allows developers to better control the development environment, give full play to their personal strengths, and improve work efficiency and concentration. This highly personalized design makes DreamView+ a more flexible and adaptable development tool, helping developers complete complex development tasks more efficiently.

In the upgrade of the new version of DreamView+, the integration of local and cloud resources management provides developers with an extremely efficient development environment, which directly affects the smoothness and efficiency of development work.

Integrated resource center: The resource center that integrates cloud and local resources allows developers to more easily obtain various resources required for autonomous driving development, such as data packages, map data, and vehicle information. Developers no longer need to manually download and manage these resources, which greatly simplifies the debugging and development process and improves work efficiency.
Reduce manual operations: After upgrading, developers do not need to manually download patches or data packages locally, which eliminates tedious manual operations. This automated resource acquisition process saves a lot of time, allowing developers to focus more on core development tasks rather than resource management.
Centralized resource management experience: By providing a convenient and centralized resource management experience, the new version of DreamView+ enables developers to manage and browse various resources more conveniently. This centralized design helps improve developers' visibility and control over resources, further improving development efficiency.
Cloud local synchronization function: The introduction of cloud and local synchronization function solves the trouble caused by data synchronization problems in the past. Developers do not need to worry about inconsistencies between local and cloud data. The synchronization function can ensure the consistency of data in different environments, reducing the burden of synchronization operations on developers.
Focus on core development work: The entire upgrade allows developers to focus more on the development of core autonomous driving applications without being distracted by cumbersome resource management and synchronization issues. This improves the smoothness and convenience of overall development and promotes the rapid development of autonomous driving applications.

The new version of DreamView+ provides developers with a more efficient and convenient development experience through technical improvements. From resource management to the introduction of synchronization functions, it is designed to reduce developers' operational burden and enable them to focus more on creative development. Work. These improvements directly translate into a more efficient autonomous driving application development process, creating a better working environment for developers.

Summary of the new version of DreamView+

Debugging process is simpler
Window layout is more flexible
Resource access is more convenient

Document platform reconstruction

In Apollo Open Platform 9.0, the comprehensive upgrade of the Document Center starts directly from the developer's perspective, providing more refined and comprehensive development support for individual developers, academic researchers, and enterprise developers.

Customized document services: The Document Center provides specially customized development documents for different types of developers through differentiated services. This allows developers to obtain information that meets their needs more directly, improving the usefulness and customizability of documents.
Detailed code comments and descriptions: The optimized documentation focuses on code management and completeness. By improving code-related comments and descriptions, it is easier for developers to understand the logic and functions when reading the code. This directly reduces the cost of understanding the code and accelerates the developer's learning and application process.
Improve development efficiency: By optimizing documentation, developers can integrate into the Apollo development environment more quickly. Detailed code comments not only reduce the learning curve, but also reduce error rates, allowing developers to use platform features more efficiently, thus improving overall development efficiency.
Smooth reading experience: The upgrade of the document center makes reading documents smoother. The clear structure and elegant layout help developers find the information they need faster, improving the developer experience.
Support a diverse developer community: By meeting the different needs of individual developers, academic researchers, and enterprise developers, the upgrade to the Documentation Center ensures that all developers can find the support and information they need in their areas of expertise. Improved platform applicability.

This upgrade of the documentation center makes it easier for developers to understand and use Apollo Open Platform 9.0, providing more powerful support for their development work. This series of technical improvements translates directly into a more efficient and enjoyable development experience, thereby increasing overall developer satisfaction.

Document platform reconstruction summary

Reduce the cost of learning and use
More convenient to operate
Reading becomes smoother
The content is more substantial.

Large-scale implementation of application scenarios

Stronger comprehensive development capabilities

Apollo open platform version 9.0 has ushered in a series of exciting technology upgrades, providing developers of autonomous driving systems with more powerful tools and functions, and significantly improving system performance in complex scenarios. .

Precise positioning technology introduced by RTK and SLAM:

Benefits: Developers can now take advantage of highly accurate vehicle positioning, especially in complex environments. The introduction of SLAM can help overcome the problem of positioning drift and provide more reliable basic positioning data.
Developer benefits: A high level of vehicle positioning accuracy provides a more reliable basis for path planning and vehicle control. This means that developers can more accurately grasp the vehicle position in different scenarios and improve the robustness of the system.

Optimization and integration of perception modules:

Benefits: Optimize lidar and camera fusion, retrain perception models, and provide more effective perception tools. The system's robustness is enhanced and its adaptability to complex scenarios is improved.
Developer benefits: More effective perception tools mean developers can more accurately understand the surrounding environment and improve the system's ability to understand complex scenes. This provides developers with more reliable perception data, helping to build smarter and more responsive autonomous driving systems.

New background segmentation algorithm and special-shaped obstacle detection:

Benefits: For special-shaped obstacles in complex environments, the new algorithm combines deep learning technology to provide a more comprehensive solution.
Developer benefits: More accurate detection of unusual obstacles means developers can more effectively avoid elusive, unconventional obstacles. This provides key advantages for system security and resilience.

Improved accuracy and smoothness of planning control:

Benefits: More flexible and stable path planning and vehicle control, providing users with a more comfortable and safer driving experience.
Developer benefits: Higher levels of control accuracy mean developers can make smarter driving decisions and provide a smoother driving experience. This gives developers greater control, allowing them to fine-tune the system to meet specific needs.

These technical upgrades in version 9.0 of the Apollo open platform provide developers with more powerful and flexible tools, allowing developers to more easily handle complex scenarios, improve system performance, and optimize user experience. By introducing advanced technologies and optimization algorithms, this version creates a more creative and development environment for developers, helping to build high-performance, reliable autonomous driving systems.

Rich scene capabilities

When dealing with the challenge of scene enrichment, the Apollo team focuses on three core dimensions to provide developers with more flexibility and convenience, allowing them to develop and deploy autonomous driving systems more efficiently.

Standard protocol models integrate specific vehicle behaviors and functions:

Technology update: Introducing a standard protocol model that allows autonomous vehicles in different scenarios to implement special operations, such as the lifting or dropping of mining trucks, and the cleaning or spraying of sanitation vehicles.
Developer benefits: Developers can easily integrate various vehicle behaviors through standardized interfaces and assign specific functions to the vehicle. This provides developers with more highly customizable solutions that can be adapted to various special application scenarios.

Flexibility in parameter allocation and interface calling:

Technical update: Introducing parameter allocation and interface calling mechanisms, allowing developers to grant more effective capabilities to vehicles based on special scenarios or operating environments, including control of speed, route, driving range, and right of way.
Developer benefits: Developers can control autonomous vehicles more precisely according to specific needs and improve the adaptability of the system. This flexibility allows developers to better meet the customized needs of different industries and application scenarios.

The introduction of closed-loop systems and third-party assistance:

Technology update: Introduce third-party assistance, involving system scheduling, remote driving, etc., to build a complete closed-loop autonomous driving application. This includes collaboration with partners to enhance the overall performance and functionality of the system.
Developer benefits: Developers can take advantage of the complete closed-loop system and easily integrate third-party assistance to improve the robustness of the application system. This integration and collaboration provides developers with a more efficient development, testing and deployment process.

These technical improvements create a more powerful and flexible development environment for developers. Developers can better respond to diverse application scenarios and industry needs, improve development efficiency, and provide stronger support for the expansion and adaptability of autonomous driving systems.

Hardware selection and operation and maintenance tools

In version 9.0, it fully supports ARM architecture compilation and can run stably and with high performance at the scene general capability layer. At the same time, a number of advanced sensors have been introduced, bringing extensive expansion to the hardware BOM map. For partners who choose a software platform for vehicle deployment, this means they have more flexible choices in terms of hardware configuration. From a developer's perspective, this series of technology upgrades and hardware expansions have brought the following significant benefits:

Comprehensive support for ARM architecture compilation:

Technology update: The introduction of ARM architecture compilation support means that developers can make full use of the advantages of the ARM architecture to achieve more efficient and flexible hardware resource management and utilization.
Developer benefits: Developers can more directly optimize and customize software to take full advantage of the performance and energy efficiency of the new generation ARM architecture. This gives developers greater flexibility to better adapt to different computing needs.

Introduction of new sensors:

Technology update: Introducing a number of new sensors to provide vehicles with richer and more accurate perception capabilities, including but not limited to advanced vision, radar and laser sensors.
Developer benefits: Developers can make fuller use of data from new sensors to improve vehicle perception and environmental understanding. This provides more possibilities for developing more complex and intelligent autonomous driving algorithms, promoting innovation and continuous evolution.

Flexibility in hardware selection:

Technology updates: Provides a wider range of hardware options, allowing partners to flexibly configure vehicle hardware based on specific needs and budget requirements.
Developer benefits: Developers can choose the most appropriate hardware configuration based on application scenarios and performance requirements to maximize the advantages of software and hardware working together and improve the overall system performance.

Recommended operation and maintenance tools:

Technical update: Two operation and maintenance tools are recommended, focusing on large-scale deployment and subsequent continuous operations to improve system stability and maintainability.
Developer benefits: The recommendation of operation and maintenance tools means that developers can use these tools to manage and maintain vehicle systems more efficiently, reduce operating costs, and focus on algorithm optimization and innovative development.

This series of technology upgrades and hardware expansions provide developers with more innovative possibilities and greater flexibility, allowing them to better adapt to evolving autonomous driving technology requirements.

Technical advantages and developer benefits of sensor calibration tools:

Visual calibration of lidar and cameras:

Technical update: The sensor calibration tool supports visual calibration of lidar and cameras, providing developers with an intuitive and accurate calibration interface.
Developer benefits: Developers can more easily perform sensor calibration and adjust parameters intuitively through the visual interface, reducing the cost of trial and error and improving calibration accuracy.

The success rate is above 90%:

Technical update: The success rate of the tool is above 90%, indicating that the calibration process is highly stable and reliable.
Developer benefits: Improved calibration success rate means developers can complete calibration work faster, reduce the number of recalibrations, and save time and resources.

Cost savings:

Technical update: The tool's high success rate and visual calibration reduce the labor and time costs required for developers to perform calibration.
Benefits to developers: The reduction in costs allows developers to invest more resources in algorithm optimization and innovation, improving overall R&D efficiency and quality.

Calibration time is shortened to hours:

Technology update: By optimizing the process and simplifying steps such as uploading data from the cloud, the entire calibration process time is shortened to hours.
Developer benefits: Reduced calibration time means developers can verify and iterate algorithms more quickly, speeding up the development cycle and improving R&D efficiency.

The technical advantages and developer benefits of the full-process tool for map collection, cartography, and editing:

Full process tool support:

Technical update: The full-process tool for map collection, mapping, and editing provides developers with an integrated solution and simplifies the process of map-related work.
Developer benefits: Developers do not need to switch between different tools, which improves work consistency and reduces learning and usage costs.

The foundation for large-scale implementation is laid:

Technology update: The tool supports fast and efficient map production, providing a foundation for the large-scale implementation of autonomous vehicles in different regions.
Developer benefits: Developers can focus more on optimizing algorithms and models without spending too much energy on the tedious process of map production, which improves development efficiency.

Faster scale implementation:

Technology update: The efficiency and integration features of the tool enable vehicles to be deployed on a larger scale in different regions more quickly.
Developer benefits: Developers can respond more flexibly to different geographical and road conditions, shortening the time from the laboratory to actual applications, and promoting the rapid advancement of projects.

These two tools provide technically efficient and reliable solutions, saving developers time and resource costs, allowing them to focus more on the core innovation of autonomous driving technology and the optimization of application scenarios.

Customized training of industry talents

Benefits of the latest developer community framework and Baidu Apollo open platform 9.0 to developers:

A comprehensive community integrating the three abilities of learning, practicing and experiencing:

Technology update: The community provides ability training in three aspects: learning, practice and experience to meet the comprehensive needs of developers in the field of autonomous driving. Learning covers theoretical knowledge, exercises provide practical opportunities, and acceptance tests apply practical abilities through competitions.
Developer benefits: Developers can improve their skills in a comprehensive learning environment, not only learning theoretical knowledge, but also exercising their abilities through practice and application testing.

Based on the underlying Apollo open platform 9.0:

Technology update: The community is built on Apollo's open platform, providing underlying technical support and resources to enable developers to more easily access autonomous driving technology.
Developer benefits: Developers can directly use the underlying Apollo platform, saving time on building a basic framework, and can focus more on their own applications and innovation.

Provide learning, hands-on experimental training and competition services:

Technical update: The community provides developers with a full range of services, including learning resources, hands-on experimental training, and opportunities to participate in various competitions.
Benefits for developers: Developers can obtain a wealth of learning materials, practice opportunities and competition experience through the community, and comprehensively improve their technical level.

Discover potential talents of autonomous driving practitioners:

Technology update: Through community learning, practice and competition testing, the Baidu team has successfully discovered many potential autonomous driving practitioners.
Benefits for developers: Developers have the opportunity to demonstrate their abilities by actively participating in community activities, and may be discovered as outstanding talents in the industry, improving personal career development opportunities.

Promote the talent training process and jointly promote scientific and technological development:

Technology update: The community has in-depth cooperation with the China Robotics and Artificial Intelligence Competition to jointly promote the training process of scientific and technological talents and contribute to the development of China's intelligent technology.
Benefits for developers: Developers can communicate with outstanding talents in the industry by participating in cooperative events, promote their own technical exchange and learning, and also have the opportunity to demonstrate their technical strength.

This comprehensive community and cooperation with the competition provide developers with a comprehensive learning, practice and competition platform, helping to cultivate autonomous driving developers with innovative capabilities and practical experience, while also promoting talent training and technology in the entire industry develop.

Benefits of Baidu Apollo Autonomous Driving Simulation Competition to developers:

Richness of practical opportunities:

Technology Update: The Apollo Autonomous Driving Simulation Competition provides developers with a wealth of practical opportunities. By participating in the competition, developers can simulate real road situations in a virtual environment to test and improve autonomous driving algorithms.
Developer benefits: Developers can accumulate experience in the simulation environment of actual scenarios, improve their understanding and operating skills of the autonomous driving system, and thus better cope with actual challenges.

Broadness of participating teams and institutions:

Technical update: More than 1,500 teams come from more than 300 colleges and universities across the country, demonstrating the broad participation of the event.
Developer benefits: Developers have the opportunity to communicate with peers from different institutions, share experiences and learning results, expand interpersonal networks, and promote cooperation and mutual growth.

Cultivating future autonomous driving talents:

Technology update: The competition helps identify and cultivate future autonomous driving talents in schools, and cultivates professionals with practical experience and technical attainments for the industry.
Developer benefits: Participants have the opportunity to hone their skills through simulation competitions, become professionals in the future autonomous driving field, and increase their employment competitiveness in this field.

Promote the popularization of autonomous driving technology:

Technology update: The number of participants of more than 5,000 people shows that the self-driving simulation competition has successfully attracted widespread attention and promoted the popularity of self-driving technology among the developer community.
Developer benefits: Developers have the opportunity to access the latest autonomous driving technology, participate in promoting industry development, and increase their personal technical reserves.

The solid foundation laid:

Technology update: The successful holding of the event laid a solid foundation for the future development of autonomous driving technology, especially remarkable results in cultivating talents in schools.
Developer benefits: Developers benefit from this successful foundation and are expected to work and innovate in a more solid technology environment in the future.

The Apollo autonomous driving simulation competition provides developers with a wealth of practical opportunities, promotes talent training and technical exchanges, and lays a solid foundation for the promotion and popularization of autonomous driving technology.

Benefits of Baidu team school-enterprise cooperation to developers:

Comprehensive and solid resource construction:

Technical update: The Baidu team has provided developers with comprehensive learning resources by building a curriculum system and a "dual-qualified" teacher training system.
Developer benefits: Developers can benefit from complete educational resources and learn basic theoretical and practical knowledge in the field of autonomous driving.

Practical opportunities for integrating industry and education:

Technology update: Through the integration of industry and education, the Baidu team has established an experimental base to provide simulation experiments and vehicle training experiments, allowing students to directly participate in technical verification.
Developer benefits: Developers can apply theoretical knowledge in actual scenarios, improve practical operation capabilities, and better adapt to the actual needs of autonomous driving technology.

A clear definition of the professional talent gap:

Technical update: Baidu team has sorted out a clear definition of the majors, training directions and talent positions for autonomous driving talent gaps.
Benefits for developers: Developers can choose their learning direction more specifically, accurately match industry needs, and increase their competitiveness in the workplace.

Rich tutorials and teaching materials:

Technology update: In the process of teacher training and course training, the Baidu team has developed rich and specialized tutorials and teaching materials for autonomous driving talent training.
Developer benefits: Developers can obtain high-quality learning materials and master the relevant knowledge and skills of autonomous driving technology more efficiently.

In-depth cooperation and successful cases:

Technology update: Baidu team has conducted in-depth cooperation with many universities and formed some successful cases and benchmarks.
Benefits for developers: By studying successful cases, developers can understand industry best practices, learn from successful experiences, and improve their career development prospects.

Strong support from the service guarantee system:

Technology update: Baidu team provides a powerful service guarantee system, including teacher training, teaching material refresh, competition introduction, etc., covering a wider range of student groups.
Developer benefits: Developers can enjoy a full range of support and services to better cope with learning and practice challenges.

The Baidu team's school-enterprise cooperation provides developers with systematic and rich learning resources, strengthens the combination of theory and practice, clarifies the direction of talent training, and provides strong support for developers' career development in the field of autonomous driving.

Use personal cases to quickly unbox DreamView+

Experiment purpose: quickly display high-precision maps into DreamView+

Experimental steps:

Create a new map file

Reference structure:

Create a new map folder under APOLLO_ROOT/modules/map/data/ and put base_map.bin in the folder.

Create your own map under modules/map/data, such as highway101

Get base_map.bin

You can find the map you want to load in the library library in LGSVL, then download the .bin map file corresponding to the apollo format and rename it to base_map.bin

If it is in other formats, such as OpenDrive files, you need to switch to the format supported by Apollo. bin file

Generate map:

base_map is the most complete map, containing all road and lane geometry and markings. Other versions of maps are generated based on base_map.
routing_map contains the topology structure of the lane in base_map, which can be generated by the following command: dir_name=modules/map/data/demo#examplemapdirectory./scripts/generate_routing_topo_graph.sh --map_dir ${dir_name}
sim_map is a lightweight version based on base_map suitable for Dreamview visual visualization. Data density has been reduced for better runtime performance. Can be generated by the following command: dir_name=modules/map/data/demo#examplemapdirectorybazel-bin/modules/map/tools/sim_map_generator --map_dir= dirname − − outputdir = {dir_name} --output_dir= dirn ame −− outputd ir={dir_name}

generate_map.sh is as follows:

#!/usr/bin/env bash

DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"

cd "${DIR}/.."

source "$DIR/apollo_base.sh"

if [ $# -eq 0 ]; then

  echo "Please specify name of your map directory."

else

  dir_name=modules/map/data/$1

  bazel-bin/modules/map/tools/sim_map_generator --map_dir=${dir_name} --output_dir=${dir_name}

  bash scripts/generate_routing_topo_graph.sh --map_dir ${dir_name}

Execute the following command:

cd /apollo

generate_map.sh YOUR_MAP_FOLDER_NAME

Restart dreamview:

./scripts/bootstrap.sh stop && ./scripts/bootstrap.sh

或

./scripts/bootstrap.sh restart

Personal steps:

First add a map from the library, such as highway101GE

Then download the corresponding base_map.bin

Then in the docker environment,

bash ./scripts/generate_routing_topo_graph.sh --map_dir modules/map/data/highway101

生成routing_map.bin routing_map.txt

bazel-bin/modules/map/tools/sim_map_generator --map_dir=modules/map/data/highway101 --output_dir=modules/map/data/highway101

Generate sim_map.bin sim_map.txt

Use of mapshow

Simply put, mapshow is a tool for visualizing high-precision maps using python scripts in the Apollo framework.

This tool is located in the "/apollo/modules/tools/mapshow" directory by default.

All you can do is run the following command to add the library to the python runtime library. Of course, you can also run it directly in this directory without adding a running environment.

source scripts/apollo_base.sh
usage: python mapshow.py [-h] -m MAP [-sl] [-l LANEID [LANEID …]]
optional arguments:
-h, --help show this help message and exit
-m MAP, --map MAP Specify the map file in txt or binary format
-sl, --showlaneids Show all lane ids in map
-l LANEID [LANEID …], --laneid LANEID [LANEID …] Show specific lane id(s) in map

Following the above steps, a brand new high-precision map was completely displayed in DreamView+ in less than ten minutes .

(This picture has been marked and blurred)

Because version 9.0 has the visualization tool DreamView+, for developers, it can be said to have a pair of wings in daily development, solving the previous problems of visualization and virtual environment deployment and scheduling.

Experience after using Apollo open platform 9.0 Beta version

I have heard of the word Apollo a few years ago, and people in the industry call it "Apollo". When I first came into contact with computer vision, I was studying a visual library such as opencv. At that time, I was thinking about when we in China would be able to implement computer vision algorithms and visualization operating platforms as well as the complete documentation platform and simulation platform behind them. "One chain" service. After I tried the 9.0 Beta version of the Apollo open platform, I thought that our mainland China had already achieved a system that transcends the world. I think the support for the new 4D millimeter wave radar will allow developers to select a cost-effective auxiliary sensor in addition to the system dominated by visual perception algorithms. Based on the original lidar point cloud processing algorithm, There is another layer of new algorithm architecture.

Regarding the construction of the Apollo Studio developer community, it can be said that we developers who love the field of computer vision have found their own home. Its original intention is to "one-stop learning and practice community", which is different from the developer community of ordinary platforms. The difference is that it implements a "one-stop" service, including: building documents/courses [Learning] - Experiments [Practice] - Competitions [Verification], the developer learning journey. I believe this is what we develop The community service system is what our readers need most and is the most helpful to us.

Regarding the release of DreamView+, this is what I am most excited about and the most shocked about the software part of version 9.0 of the Apollo open platform. I think this is an epic update for developers in the field of manned autonomous driving. Before this, there was no HMI interaction system that could be highly customized, so developers had to combine the smart driving domain with the smart cockpit domain. To implement split design, while increasing the system complexity, it is also necessary to maintain a complete set of system software communication platforms to support the commercial application of intelligent driving systems and intelligent interactive systems. The release of DreamView+ allows developers to experience the "cabin-driving integrated" technical implementation solution directly out of the box without having to go through the complicated steps before, allowing developers to gain a pair of wings during the development process and enter. Development of the “green channel” model.

In short, the Apollo open platform version 9.0 updated by the Baidu Apollo team has given me a lot of surprises. These surprises are what our developers want, what Chinese companies want, and what the global industry wants!

my expectations

I hope that many developers like me who are passionate about computer vision can join Baidu's Apollo developer community, learn and progress together with many people who share the same passion, and elevate China's autonomous driving field to a higher level.
I hope that the Baidu Apollo team can continue to try better innovations and innovate better iconic products of the era while better maintaining the developer community and other technologies, because in the eyes of the new era, there is no best, only better. .
It is hoped that enterprises led by new industries can maintain a win-win mentality and work with cutting-edge technology teams like Xiang Baidu Apollo team to create industries belonging to the new era.

If my expectations can bring better progress to the next innovation and practice of the Baidu Apollo team, many developers and new industry-led enterprises, I will be deeply honored!