How autowarefoundation/autoware_core works
Main Components
• Autoware Core Entry Point and Launch Infrastructure: The top-level “autoware_core” folder (with its CMakeLists.txt, launch files, and package.xml) is the orchestrator.
• Modular ROS Packages and Libraries: The repository is divided into key subsystems, typically structured into directories such as:
– common/ (for shared utilities and component interface specifications)
– control/ (for vehicle control and pure pursuit drivers)
– localization/ (for various localization nodes, including EKF and stop filter modules)
– map/ (for map loading, projection, and lanelet2 interfaces)
– perception/ (for ground filtering and core perception modules)
– planning/ (for mission planning, trajectory generation, and velocity smoother components)
– vehicle/ (for vehicle-specific drivers and URDF definitions)
• Additional Modules: There are packages for testing, simulation (e.g., autoware_testing), and experimental visualization (e.g., autoware_pyplot).
Relationships and Interactions
• Library Dependencies and Shared Utilities: The “common” directory provides lower-level utilities (such as interpolation, Kalman filters, and motion utils) that are used by multiple subsystems. Represent these as shared libraries that “support” the higher-level packages.
• Launch and Parameter Management: Each module contains launch files (often in XML or Python) and parameter configuration (YAML files) that bind the system together – highlighting that configuration and runtime initialization are essential parts of the overall design.
Key Architectural Patterns
• Modularity Through ROS Packages: The design pattern here is modular, with each package serving as a self-contained node or set of nodes following ROS best practices.
• Abstraction of Component Interfaces: “common/autoware_component_interface_specs” formalizes interfaces between components, allowing decoupling and easier extension.
• Distributed and Layered System: Sensors, perception, planning, and control are implemented as separate nodes communicating over ROS middleware – a pattern standard in autonomous driving architectures.
Relevant Technologies and Frameworks
• ROS (Robot Operating System) as the underpinning middleware.
• C++ as the predominant language in the packages, with CMake for builds.
• Configuration using YAML (for parameters) and XML/launch files to bootstrap node execution.
• Additional tools (e.g., OSQP for optimization interfaces, lanelet2, and third-party visualization utilities).