Determining component dependencies#
For any developers who wish to try and deploy Autoware as a microservices architecture, it is necessary to understand the software dependencies, communication, and implemented features of each ROS package/node.
As an example, the commands necessary to determine the dependencies for the Perception component are shown below. In Autoware 2.0.0, the launch files of the Perception component are in the autoware_perception_launch package. This package is in the autoware_launch repository.
Run the commands of this page from the root of your Autoware workspace, for example ~/autoware.
Perception component dependencies#
Package dependencies#
To generate a graph of package dependencies, use the following colcon command. The dot command comes from the graphviz package (sudo apt install graphviz).
colcon graph --dot --legend --packages-up-to autoware_perception_launch | dot -Tpng -o graph.png
In the graph, a blue edge is a build dependency, a red edge is a run dependency, a tan edge is a test dependency, and a dashed edge is an indirect dependency. The following diagram groups the packages of this graph for Autoware 2.0.0.
The diagram does not show which package depends on which. To see each dependency, open the graph.png file from the colcon graph command.
To generate a list of dependencies, use:
colcon list --packages-up-to autoware_perception_launch --names-only
colcon list output (Autoware 2.0.0, 102 packages)
autoware_adapi_v1_msgs
autoware_agnocast_wrapper
autoware_bevfusion
autoware_bytetrack
autoware_cmake
autoware_common_msgs
autoware_compare_map_segmentation
autoware_crosswalk_traffic_light_estimator
autoware_cuda_dependency_meta
autoware_cuda_pointcloud_preprocessor
autoware_cuda_utils
autoware_detected_object_feature_remover
autoware_detected_object_validation
autoware_detection_by_tracker
autoware_elevation_map_loader
autoware_euclidean_cluster
autoware_grid_map_utils
autoware_ground_segmentation
autoware_ground_segmentation_cuda
autoware_image_object_locator
autoware_image_projection_based_fusion
autoware_image_transport_decompressor
autoware_internal_debug_msgs
autoware_internal_localization_msgs
autoware_internal_metric_msgs
autoware_internal_msgs
autoware_internal_perception_msgs
autoware_internal_planning_msgs
autoware_interpolation
autoware_kalman_filter
autoware_lanelet2_extension
autoware_lanelet2_utils
autoware_lidar_apollo_instance_segmentation
autoware_lidar_centerpoint
autoware_lidar_transfusion
autoware_lint_common
autoware_map_based_prediction
autoware_map_msgs
autoware_motion_utils
autoware_multi_object_tracker
autoware_object_merger
autoware_object_recognition_utils
autoware_object_sorter
autoware_occupancy_grid_map_outlier_filter
autoware_pcl_extensions
autoware_perception_config
autoware_perception_launch
autoware_perception_msgs
autoware_planning_msgs
autoware_point_types
autoware_pointcloud_preprocessor
autoware_probabilistic_occupancy_grid_map
autoware_ptv3
autoware_pyplot
autoware_qos_utils
autoware_raindrop_cluster_filter
autoware_sensing_msgs
autoware_shape_estimation
autoware_simpl_prediction
autoware_simple_object_merger
autoware_tensorrt_bevdet
autoware_tensorrt_classifier
autoware_tensorrt_common
autoware_tensorrt_plugins
autoware_tensorrt_yolox
autoware_test_utils
autoware_tracking_object_merger
autoware_traffic_light_arbiter
autoware_traffic_light_category_merger
autoware_traffic_light_classifier
autoware_traffic_light_fine_detector
autoware_traffic_light_map_based_detector
autoware_traffic_light_multi_camera_fusion
autoware_traffic_light_occlusion_predictor
autoware_traffic_light_selector
autoware_traffic_light_utils
autoware_traffic_light_visualization
autoware_utils
autoware_utils_debug
autoware_utils_diagnostics
autoware_utils_geometry
autoware_utils_logging
autoware_utils_math
autoware_utils_pcl
autoware_utils_rclcpp
autoware_utils_system
autoware_utils_tf
autoware_utils_uuid
autoware_utils_visualization
autoware_vehicle_info_utils
autoware_vehicle_msgs
bevdet_vendor
cuda_blackboard
managed_transform_buffer
mussp
negotiated
negotiated_interfaces
perception_utils
pointcloud_to_laserscan
tier4_debug_msgs
tier4_external_api_msgs
tier4_perception_msgs
Tip
To output a list of modules with their respective paths, run the command above without the --names-only parameter.
System dependencies#
The packages also depend on ROS packages and system libraries that are not in the Autoware workspace. rosdep lists them as rosdep keys. To list the keys of the whole Perception component, give rosdep the paths of all packages from the colcon list command:
rosdep keys --ignore-src --from-paths $(colcon list --packages-up-to autoware_perception_launch --paths-only)
The --ignore-src option removes the packages that are in the given paths. The output contains keys such as libpcl-all-dev, libopencv-dev, tensorrt_cmake_module, lanelet2_core, and grid_map_ros. To see the system package for one key, use rosdep resolve:
rosdep resolve libpcl-all-dev
Unused dependencies#
A package.xml file can declare a dependency that the code does not use. The autoware_dependency_checker package of autoware_tools compares the dependencies in each package.xml file with the headers that the source files include. It mainly checks the packages whose names start with autoware_. To get it, import repositories/tools.repos into your workspace and build it:
vcs import src < repositories/tools.repos
colcon build --packages-up-to autoware_dependency_checker
source install/setup.bash
Then run the script in a package folder or in the workspace root:
cd src/universe/autoware_universe/perception/autoware_lidar_centerpoint
ros2 run autoware_dependency_checker dependency_checker.sh
Perception component communication#
To see which ROS topics are being subscribed and published to, launch Autoware and run rqt_graph. For example, start the rosbag replay simulation, and then run:
ros2 run rqt_graph rqt_graph
In rqt_graph, enter /perception in the namespace filter to show only the nodes and topics of the Perception component.