apollo-v9.0.0

Apollo Open Source Platform 9.0 further focuses on enhancing the development and debugging experience, dedicated to providing autonomous driving developers with a unified development tool platform and easy-to-extend PnC and perception software framework interfaces. The new version reshapes the PnC and perception extension development method based on package management. It optimizes component splitting and configuration management according to business logic, simplifying the process of calling. In addition to the component extension method, a more lightweight plugin extension method has been added, simplifying the process of extending. The new version introduces Dreamview Plus, a brand-new developer tool that introduces modes for convenient multi-scenario use, a panel layout for customizing visualization, and a resource center providing richer development resources. Furthermore, the LiDAR and Camera detection models in the new version have been upgraded for improved results, and incremental training methods have been opened up for easy extension. At the same time, support for 4D millimeter-wave radar has been added. Finally, the new version is adapted to the ARM architecture, and supports compilation and running on Orin, providing developers with additional device options.

Major New Features and Improvements:

• PnC Extension Development Pattern Based On Package Management

  • Unified external interfaces to decouple the operation layer and PNC module.
  • A brand-new plugin extension method to facilitate developers in developing and deploying their functionality.
  • Global parameters and local parameters are divided to allow developers to query and modify parameters.

• Perception Extension Development Pattern Based On Package Management

  • Re-split the perception components based on "functional" granularity to facilitate reuse.
  • A brand-new plugin development mode to facilitate the replacement of algorithms under the existing perception pipeline.
  • Simplified and unified configuration to allow developers to query and modify parameters at any time.

• Brand New Dreamview Plus Developer Tool

  • Organize the usage scenarios of the development tool based on "mode", such as Perception mode, PnC mode, and Vehicle Test mode.
  • Encapsulate each visualization functionality into an independent panel and support developers to customize the panel.
  • Provide various resources such as maps, scenarios, vehicle configurations, and data records in the resource center to facilitate development and debugging.

• Fully Upgraded Perception Model to Support Incremental Training

  • CenterPoint is adopted to replace CNNSeg model by default in the LiDAR perception pipeline and YOLOX+YOLO3D is adopted to replace the original YOLO model by default in the Camera perception pipeline.
  • Provide incremental training to improve perception model capabilities in specific scenarios by using a small amount of annotated data and Apollo pre-trained models.
  • Support 4D millimeter-wave radar from hardware driver to perception model layer.

[Note] All models and methodologies included in Apollo Open Source Platform 9.0 are for research purposes only. Productized and commercialized uses of these models are NOT encouraged, and it is at your own risk. Please be cautious to try Apollo Open Source Platform 9.0 with sufficient safety protection mechanisms in place. Your feedback is highly appreciated so that we can continuously improve the models.

apollo-v8.0.0

Apollo 8.0 is an effort to provide an extensible software framework and complete development cycle for Autonomous Driving developer. Apollo 8.0 introduces easily-reused Package to organize software modules. Apollo 8.0 integrates the whole process of perception development ,by combining model training service, model deployment tool and end-to-end visual validation tool . And another 3 new deep learning models are incorporated in Apollo 8.0 for perception module. Simulation service is upgraded by integrating local simulator in Dreamview to provide powerful debug tool for PnC developer.

Major Features and Improvements

• Reusable Software Package
  • Reorganize the modules based on Package to provide the functionality in an easy-to-consume manner
  • Fast installation experience based on Package, refer to Installation - Package Method
  • Support customizing and sharing Package
• Brand New Deep Learning Models
  • CenterPoint, center-based two-stage 3D obstacle detection model
  • CaDDN, camera obstacle detection model
  • BEV PETR, camera obstacle detection model
• Complete Perception Development Process
  • Support Paddle3D to provide Model Training service
  • Provide model deployment tool by normalizing the model meta.
  • Provide visual validation tool in Dreamview
• Upgraded PnC Simulation Service
  • Provide PnC debug tool by integrating local simulator in Dreamview
  • Support scenario editing online and download in Dreamview

[Note] All models and methodologies included in Apollo 8.0 are for research purposes only. Productized and commercial uses of these models are NOT encouraged, and it is at your own risk. Please be cautious to try Apollo 8.0 with enough safety protection mechanism. Your feedback is highly appreciated for us to continuously improve the models.

apollo-v7.0.0

Apollo 7.0 incorporates 3 brand new deep learning models to enhance the capabilities for Apollo Perception and Prediction modules. Apollo Studio is introduced in this version, combining with Data Pipeline, to provide a one-stop online development platform to better serve Apollo developers. Apollo 7.0 also publishes the PnC reinforcement learning model training and simulation evaluation service based on previous simulation service.

Major Features and Improvements

• Brand New Deep Learning Models
  • Mask-Pillars obstacle detection model based on PointPillars
  • Inter-TNT prediction model based on interactive prediction & planning evaluator
  • Camera obstacle detection model based on SMOKE
• Apollo Studio Services
  • Practice environment service
  • Vehicle management service
• PnC Reinforcement Learning Services
  • Smart training and evaluation service
  • Extension interface
• Upgraded Perception Module Code Structure

[Note] All models and methodologies included in Apollo 7.0 are for research purposes only. Productized and commercial uses of these models are NOT encouraged, and it is at your own risk. Please be cautious to try Apollo 7.0 with enough safety protection mechanism. Your feedback is highly appreciated for us to continuously improve the models.

apollo-v6.0.0

Apollo 6.0 incorporates new deep learning models to enhance the capabilities for certain Apollo modules. This version works seamlessly with new additions of data pipeline services to better serve Apollo developers. Apollo 6.0 is also the first version to integrate certain features as a demonstration of our continuous exploration and experimentation efforts towards driverless technology.

Major Features and Improvements

• Upgraded Deep Learning Models
  • PointPillars based obstacle detection model
  • Semantic map based pedestrian prediction model
  • Learning based trajectory planning model
• Brand New Data Pipeline Services
  • Low speed obstacle prediction model training service with semantic map support
  • PointPillars based obstacle detection model training service
  • Control profiling service
  • Vehicle dynamic model training service
  • Open space planner profiling service
  • Complete control parameter auto-tune service
• Driverless Research
  • Remote control interface with DreamView integration
  • Audio based emergency vehicle detection system
• Upgraded dev environment including build and dependency updates

[Note] All models and methodologies included in Apollo 6.0 are for research purposes only. Productized and commercial uses of these models are NOT encouraged, and it is at your own risk. Please be cautious to try Apollo 6.0 with enough safety protection mechanism. Your feedback is highly appreciated for us to continuously improve the models.

apollo-v5.5.0

Apollo 5.5 enhances the complex urban road autonomous driving capabilities of previous Apollo releases, by introducing curb-to-curb driving support. With this new addition, Apollo is now a leap closer to fully autonomous urban road driving.
The car has complete 360-degree visibility, along with upgraded perception deep learning model a brand new prediction model to handle the changing conditions of complex road and junction scenarios, making the car more secure and aware. New Planning scenarios have been introduced to support curb-side functionality.

Major Features And Improvements

• Brand new Data Pipeline Service
  • Sensor Calibration service
• Brand new module - Storytelling
• Scenario - Based Planning with a new planning scenarios to support curb-to-curb driving
  • Park-and-go
  • Emergency
• Prediction Model - Caution Obstacle
  • Semantic LSTM evaluator
  • Extrapolation predictor
• Control module
  • Model Reference Adaptive Control (MRAC)
  • Control profiling service
• Simulation scenarios

Autonomous Drive Capabilities

Vehicles with this version can drive autonomously in complex urban road conditions including both residential and downtown areas. BE CAUTIOUS WHEN DRIVING AUTONOMOUSLY, ESPECIALLY AT NIGHT OR IN POOR VISION ENVIRONMENT. URBAN DRIVING INVOLVES NAVIGATING HIGH RISK ZONES LIKE SCHOOLS, PLEASE TEST APOLLO 5.5 WITH THE SUPPORT FROM APOLLO ENGINEERING TEAM, PLEASE AVOID DRIVING THE VEHICLE ON THE HIGHWAY OR AT SPEEDS THAT ARE ABOVE OUR SUPPORTED THRESHOLD.

apollo-v5.0.0

Apollo 5.0 is an effort to support volume production for Geo-Fenced Autonomous Driving. The car now has 360-degree visibility, along with upgraded perception deep learning model to handle the changing conditions of complex road scenarios, making the car more secure and aware. Scenario-based planning has been enhanced to support additional scenarios like pull over and crossing bare intersections.

Major Features And Improvements

• Brand new Data Pipeline Service
  • Vehicle Calibration
• New Perception algorithms
• Sensor Calibration Service
• Scenario - Based Planning with a new planning algorithm, Open Space Planner and new scenarios supported
  • Intersection - STOP Sign, Traffic Light, Bare Intersection
  • Park - Valet, Pull Over
• Map Data Verification tool
• Prediction Evaluators
• Simulation web platform - Dreamland
  • Scenario Editor
  • Control-in-loop Simulation
• Cyber RT runtime framework
  • Official ARM CPU support with full docker dev environment support: https://github.com/ApolloAuto/apollo/blob/master/docs/cyber/CyberRT_Docker.md
  • Python language support for full set of Cyber RT API
  • Cyber RT API website goes online: https://cyber-rt.readthedocs.io/en/latest/

Autonomous Drive Capabilities

Vehicles with this version can drive autonomously in complex urban road conditions including both residential and downtown areas. BE CAUTIOUS WHEN DRIVING AUTONOMOUSLY, ESPECIALLY AT NIGHT OR IN POOR VISION ENVIRONMENT. URBAN DRIVING INVOLVES NAVIGATING HIGH RISK ZONES LIKE SCHOOLS, PLEASE TEST APOLLO 5.0 WITH THE SUPPORT FROM APOLLO ENGINEERING TEAM, PLEASE AVOID DRIVING THE VEHICLE ON THE HIGHWAY OR AT SPEEDS THAT ARE ABOVE OUR SUPPORTED THRESHOLD.

apollo-v3.5.0

Apollo 3.5 is capable of navigating through complex driving scenarios such as residential and downtown areas. With 360-degree visibility and upgraded perception algorithms to handle the changing conditions of urban roads, the car is more secure and aware.

Major Features And Improvements

• Upgraded Sensor Suite
  • VLS-128 Line LiDAR
  • FPD-Link Cameras
  • Continental long-range radars
  • Apollo Extension Unit (AXU)
  • Additional IPC
• Brand New Runtime Framework - Apollo Cyber RT which is specifically targeted towards autonomous driving
• New Perception algorithms
• Scenario - Based Planning with a new planning algorithm, Open Space Planner
• New Localization algorithm
• V2X Capabilities
• Open Vehicle Certification platform - 2 new vehicles added GAC GE3 and GWM WEY VV6

Autonomous Drive Capabilities

Vehicles with this version can drive autonomously in complex urban road conditions including both residential and downtown areas. BE CAUTIOUS WHEN DRIVING AUTONOMOUSLY, ESPECIALLY AT NIGHT OR IN POOR VISION ENVIRONMENT. URBAN DRIVING INVOLVES NAVIGATING HIGH RISK ZONES LIKE SCHOOLS, PLEASE TEST APOLLO 3.5 WITH THE SUPPORT FROM APOLLO ENGINEERING TEAM.

apollo-v3.0.0

Apollo 3.0 enables L4 product level solution that allows vehicles to drive in a closed venue setting at a low speed. Automakers can now leverage this one stop solution for autonomous driving without having to customize on their own.

Major Features And Improvements

• New Safety module called Guardian
• Enhanced Surveillance module - Monitor
• Hardware service layer that will now act like a platform and not a product, giving developers the flexibility to integrate their own Hardware
• Apollo Sensor Unit (ASU)
• New Gatekeeper - Ultrasonic Sensor
• Perception module changes:
  • CIPV detection/ Tailgating – moving within a single lane
  • Whole lane line support - bold line support for long range accuracy. There are 2 different types on installations for Camera, low and high installation.
  • Online pose estimation – determines angle change and estimates it when there are bumps or slopes to ensure that the sensors move with the car and the angle/pose changes accordingly
  • Visual localization – we now use camera for localization. This functionality is currently being tested.
  • Ultrasonic Sensor – Currently being tested as the final gatekeeper to be used in conjunction with Guardian for Automated Emergency brake and vertical/perpendicular parking.

apollo-v2.5.0

This release allows the vehicle to autonomously run on geo-fenced highways. Vehicles are able to do lane keeping cruise and avoid collisions with the leading vehicles.

Major Features And Improvements

• Upgrade MSF localization
• Upgrade DreamView with more visualization features
• Add HD map data collection tool
• Add vision based perception with obstacle and lane mark detections
• Add relative map to support ACC and lane keeping for planning and control
• Make dockerfile available

Autonomous Drive Capabilities

Vehicles with this version can drive autonomously on highways at higher speed without HD map support. The highway needs to have clear white painted lane marks with minimum curvatures. The performance of vision based perception will degrade significantly at night or with strong light flares. BE CAUTIOUS WHEN DRIVING AUTONOMOUSLY, ESPECIALLY AT NIGHT OR IN POOR VISION ENVIRONMENT.

apollo-v2.0.0

Apollo 2.0 enables your vehicle to drive on simple urban roads autonomously. It is able to cruise, avoid collisions with obstacles, stop at traffic lights and change lanes.

Read our Release Note(https://github.com/ApolloAuto/apollo/blob/master/RELEASE.md) to learn more about major features and improvements. Open a new issue(https://github.com/ApolloAuto/apollo/issues) if you have questions and feedback. And of course, we always appreciate your contribution (https://github.com/ApolloAuto/apollo/blob/master/CONTRIBUTING.md).

You can get started with Apollo 2.0(https://github.com/ApolloAuto/apollo/archive/v2.0.0.tar.gz) today.