autowarefoundation / autoware-projects

This repository is made to keep track of Non-source code related tasks for AWF projects
8 stars 0 forks source link

Data recording minimum requirements #25

Open ghost opened 2 years ago

ghost commented 2 years ago

Checklist

Description

List the minimum set of ROS topics that should be recorded in order to enable reproduction of errors that occur during field testing

Purpose

In order to fix Autoware errors or failures that occur during field testing, a minimum data set must be defined. Doing so will enable developers to:

A minimum data set will also provide a foundation for developing functionality to automatically record data when field testing on time or maximum file size requirement (eg: record for one hour or until 1TB of storage is used up, then overwrite the previous ROSBAG file).

Possible approaches

Although in theory, it should be possible to reproduce an error by recording all raw sensor and vehicle data, ROS 2 is not deterministic and so this is not guaranteed. One possible approach is to only replay processed input data into the specific part of the pipeline where an error has occured.

eg: to try and reproduce a planning error, only replay the relevant input data into the Planning module

Definition of done

Blocked by

ghost commented 2 years ago

@mitsudome-r Would it be possible to provide any links related to the interface discussion so that we can track progress of that to better understand when work on this issue can continue?

mitsudome-r commented 2 years ago

@LalithVipulananthan Here are the minutes from the Architecture WG discussions: https://github.com/autowarefoundation/autoware/discussions?discussions_q=label%3Aarchitecture_wg

Although in theory, it should be possible to reproduce an error by recording all raw sensor and vehicle data, ROS 2 is not deterministic and so this is not guaranteed. One possible approach is to only replay processed input data into the specific part of the pipeline where an error has occured.

At least we can start with minimal raw sensor data, and other primary data. This includes the following topics: Source topic name message type
Lidar /sensing/lidar/{lidar_name}/pointcloud_raw sensor_msgs/PointCloud2
Camera /sensing/camera/{camera_name}/image_raw sensor_msgs/Image
Camera /sensing/camera/{camera_name}/camera_info sensor_msgs/CameraInfo
GNSS /sensing/gnss/{gnss_name}/nav_sat_fix sensor_msgs/NavSatFix
GNSS /sensing/gnss/{gnss_name}/orientation autoware_sensing_msgs/GnssInsOrientationStamped.msg
GNSS /sensing/gnss/{gnss_name}/twist geometry_msgs/TwistWithCovarianceStamped
GNSS /sensing/gnss/{gnss_name}/acceleration geometry_msgs/AccelWithCovarianceStamped
Vehicle /vehicle/status/control_mode autoware_auto_vehicle_msgs/ControlModeReport
Vehicle /vehicle/status/steering_status autoware_auto_vehicle_msgs/SteeringReport
Vehicle /vehicle/status/velocity_status autoware_auto_vehicle_msgs/VelocityReport
- /tf
- /static_tf
- /map/pointcloud_map sensor_msgs/PointCloud2
- /map/vector_map autoware_auto_mapping_msgs/msg/HADMapBin
Some of the intermediate topics could include: topic name message type remark
/localization/kinematic_state nav_msgs/Odometry final position/velocity output
/localization/acceleration geometry_msgs/msg/AccelWithCovarianceStamped final accelaration output
/localization/acceleration geometry_msgs/msg/AccelWithCovarianceStamped final accelaration output
/perception/object_recognition/objects autoware_auto_perception_msgs/msg/PredictedObjects object recognition output
/perception/traffic_light_recognition/traffic_signals autoware_auto_perception_msgs/msg/TrafficSignalArray traffic light recognition result
/planning/mission_planning/route autoware_auto_planning_msgs/msg/HADMapRoute planned route
/planning/scenario_planning/trajectory autoware_auto_planning_msgs/msg/Trajectory planned trajectory
/control/command/control_cmd autoware_auto_control_msgs/msg/AckermannControlCommand control command output
stale[bot] commented 2 years ago

This pull request has been automatically marked as stale because it has not had recent activity.