Although the planning succeeds and there is no obstacle, path is replanned during driving.

[TakumIto]

Checklist

• [X] I've read the contribution guidelines.
• [X] I've searched other issues and no duplicate issues were found.
• [X] I'm convinced that this is not my fault but a bug.

Description

Although the planning succeeds and there is no obstacle, path is replanned during driving.

Expected behavior

The vehicle follows the initially planned path without replanning.

Actual behavior

The path is replanned though there is no obstacle.

Screencast from 08-09-2023 04:32:44 PM.webm

Steps to reproduce

1. Start autoware planning simulator using default. (select astar algorithms)
2. Set start and goal position using the commands below.

   ros2 topic pub --once /planning/mission_planning/goal geometry_msgs/msg/PoseStamped "
   header:
   stamp:
   sec: 0
   nanosec: 0
   frame_id: map
   pose:
   position:
   x: 3746.88
   y: 73748.5
   z: 0.0
   orientation:
   x: 0.0
   y: 0.0
   z: 0.847164
   w: 0.531332"

   ros2 topic pub --once /initialpose geometry_msgs/msg/PoseWithCovarianceStamped "
   header:
   stamp:
   sec: 0
   nanosec: 0
   frame_id: map
   pose:
   pose:
   position:
     x: 3717.34228515625
     y: 73726.71875
     z: 0.0
   orientation:
     x: 0.0
     y: 0.0
     z: 0.8393470173368638
     w: 0.5435959754152992
   covariance:
   - 0.25
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.25
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.0
   - 0.06853891909122467"

Versions

• OS: Ubuntu 22.04
• ROS2: Humble
• Autoware: Autoware.universe:main
• Map: lanelet2_map.zip

Possible causes

I think this issue is caused by the quantization error in transforming from position to index.

Additional context

No response

[VRichardJP]

I have not check in details how the parking planner is implemented, but I notice your example the vehicle is driving very close from the drivable area boundaries, which is the same than driving very close from an obstacle.

As you can see, the drivable area is updated constantly. Even if it does not change much,the changes might be enough so that a path that was considered safe is not acceptable anymore. In such case, Autoware has no choice but to generate a new path, which might be rejected 1 second later for the very same reason.

Rather than a quantization, I think this problem is inherent to A*. In particular because the algorithm does not use any soft constraint: a path that goes 1mm close from a wall is as valid as a path that stays far from every obstacle.

[stale[bot]]

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