fix(multi_object_tracker): prevent too large object tracking

[technolojin]

Description

A bug was reported that a huge (length of 6km) object is appeared from the object tracking.

It is happening when a unknown object (shape of autoware_auto_perception_msgs::msg::Shape::POLYGON) is assigned to a bicycle tracker.
It was caused by utils::convertConvexHullToBoundingBox(object, bbox_object) which treats the footprint points as global (map coordinate) position, while the input is local (the object coordinate).

1. Fix bug on utils::convertConvexHullToBoundingBox(object, bbox_object)
2. Add a detected object size checker. Update the object size only if the size is proper

Tests performed

1. Configure data_association_matrix.param.yaml to enable the unknown object can be assigned to bicycle tracker

/**:
  ros__parameters:
    can_assign_matrix:
      #UNKNOWN, CAR, TRUCK, BUS,  TRAILER, MOTORBIKE, BICYCLE,PEDESTRIAN <-Measurement
      [1,       0,   0,     0,    0,       0,         0,      0,         #UNKNOWN <-Tracker
       0,       1,   1,     1,    1,       0,         0,      0,         #CAR
       0,       1,   1,     1,    1,       0,         0,      0,         #TRUCK
       0,       1,   1,     1,    1,       0,         0,      0,         #BUS
       0,       1,   1,     1,    1,       0,         0,      0,         #TRAILER
       1,       0,   0,     0,    0,       1,         1,      1,         #MOTORBIKE
       1,       0,   0,     0,    0,       1,         1,      1,         #BICYCLE
       1,       0,   0,     0,    0,       1,         1,      1]         #PEDESTRIAN

Replay a file that contains to associate an object that have poligon shape to a tracker.

simplescreenrecorder-2024-05-29_16.56.45.mp4

Red box: previous tracking object
Green box: fixed tracking object

2. Perception Regular Test
   test result: TIER IV INTERNAL
   261/264 PASS
   ! This test shows that the reference mode can be used as before, not testing the problem is resolved (since this scene is not in the test)

Effects on system behavior

Not applicable.

Interface changes

Not applicable.

Pre-review checklist for the PR author

The PR author must check the checkboxes below when creating the PR.

• I've confirmed the contribution guidelines.
• The PR follows the pull request guidelines.

In-review checklist for the PR reviewers

The PR reviewers must check the checkboxes below before approval.

• The PR follows the pull request guidelines.

Post-review checklist for the PR author

The PR author must check the checkboxes below before merging.

• There are no open discussions or they are tracked via tickets.

After all checkboxes are checked, anyone who has write access can merge the PR.

[YoshiRi]

[imho]
I think footprint is calculated by relative distance from centroid in base_link and not in object coordinate.
So, I guess yaw rotation fix will be necessary before get maximum x,y value.

Currently, it won't make any change because yaw of unknown object may be 0 in all case.

[technolojin]

Base line: the source of the unknown polygon object is from shape_estimator.

1. Object Position:
   • Coordinate origin: base_link
   • Position: polygon centroid
   • evidence
2. Footprint:
   • Coordinate origin: object position (polygon centroid)
   • Positions: convex hull point positions, local
   • evidence

I also checked the object size, output of this function, and confirmed that the numbers are as expected (about 0.7 m).

Therefore the footprint points are not from the base_link, but from the object origin.

Oh, you are talking about the orientation... let me check it again.

[technolojin]

If the message defines the footprint is in the object coordinate (as its position is from the object position), the width and length should be on its coordinate, not rotated one.

It doesn't matter whether the shape_estimator implementation is done in yaw = 0 or not.
If the shape_estimator do the yaw estimation in the polygon&convex hull function, it should rotate the point cloud when it make the message.

[YoshiRi]

Understood. Then, your code is fine.

[YoshiRi]

LGTM