Purpose / Use cases

This package provides a ROS2 node for the computational package recordreplay_planner. This is done to separate the computation part from the ROS2-specific part. See the computational part for a rationale of what the functionality is to be used for.

Design

This is a wrapper around recordreplay_planner. Its behavior can be controlled via actions. It can record the ego state of the vehicle and play back a set of recorded states at a later time. It does stop for obstacles, from BoundingBoxArray, publsihed in any tf frame that can transform to VehicleKinematicState's frame.

Assumptions / Known limits

This node assumes that the recorded states can be played back without any tf changes. There is an open issue in https://gitlab.com/autowarefoundation/autoware.auto/AutowareAuto/issues/265 to address this limitation.

See the recordreplay_planner design documentation for more details.

Inputs / Outputs / API

Actions:

This node uses two actions to control its behavior:

• RecordTrajectory.action is used to record a trajectory. It runs until canceled. While the action is running, the node subscribes to a VehicleKinematicState.msg topic by a provided name and records all states that are published on that topic.
• ReplayTrajectory.action is used to replay a trajectory. It runs until canceled. While the action is running, the node subscribes to the same VehicleKinematicState.msg topic as when recording. When messages are published on that topic, the node publishes a trajectory starting approximately at that point (see the recordreplay_planner design documentation on how that point is determined).

The actions are defined in a separate package, recordreplay_planner_actions.

Inputs:

• autoware_auto_msgs/msg/VehicleKinematicState is the state used as recorded points for replaym, and also to prune starting point of replay trajectory
• autoware_auto_msgs/msg/BoundingBoxArray is list of bounding boxes of obstacle, that thre reaply trajectory stops for to avoid collision.

Outputs:

• autoware_auto_msgs/msg/Trajectory is the trajectory that gets published

Complexity

See recordreplay_planner.

Security considerations

TBD by a security specialist.

Future extensions / Unimplemented parts

• Proper tf support

Related issues

• #265 Extension to support tf properly
• #273 Fix and extend integration tests