feat(autoware_pointcloud_preprocessor): reuse collectors to reduce creation of collector and timer

[vividf]

Description

#8300 introduces a collector and initializes a timer in each collector with a 100 ms interval, which leads to infinite object creation and deletion.

This PR reuses collectors that are no longer in use (i.e., those that have completed concatenation) by resetting their timers, effectively converting them into new collectors. This approach helps limit unnecessary collector creation and timer initialization.

Related links

Parent Issue:

• Link

How was this PR tested?

• Test with unit test.
• Test with rosbag that provided in feat(autoware_pointcloud_preprocessor): redesign concatenate and time sync node #8300
• Test with CARET and proved that it works better than before TIER4_INTERNAL_LINK

Notes for reviewers

None.

Interface changes

None.

Effects on system behavior

None.

[github-actions]

Thank you for contributing to the Autoware project!

🚧 If your pull request is in progress, switch it to draft mode.

Please ensure:

• You've checked our contribution guidelines.
• Your PR follows our pull request guidelines.
• All required CI checks pass before marking the PR ready for review.

[codecov]

Codecov Report

Attention: Patch coverage is 56.14035% with 25 lines in your changes missing coverage. Please review.

Project coverage is 26.76%. Comparing base (4a77f3d) to head (31e99b3).
Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
...oncatenate_data/concatenate_and_time_sync_node.cpp	51.42%	12 Missing and 5 partials ⚠️
...processor/src/concatenate_data/cloud_collector.cpp	65.00%	2 Missing and 5 partials ⚠️
...sor/src/concatenate_data/combine_cloud_handler.cpp	50.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #10074      +/-   ##
==========================================
- Coverage   26.77%   26.76%   -0.01%     
==========================================
  Files        1416     1416              
  Lines      108468   108503      +35     
  Branches    41719    41730      +11     
==========================================
  Hits        29039    29039              
- Misses      76552    76587      +35     
  Partials     2877     2877              

Flag	Coverage Δ		*Carryforward flag
differential	21.15% <56.14%> (?)
differential-cuda	19.42% <56.14%> (?)
total	26.74% <ø> (-0.03%)	⬇️	Carriedforward from 887c6b7

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[YoshiRi]

LGTM in codespace, but I did not check with any rosbags.

[mojomex]

Good stuff!
I've left a few comments in places I think could be simplified or where I'd like the more safety guarantees.

[mojomex]

Probably hard-coding number of collectors to 2 or 3 would be okay. I don't see a use case with more collectors at the moment.
Also would allow us to allocate all of them once on startup instead of dynamically.

[vividf]

@mojomex @drwnz kindly ping

[drwnz]

Please see the minor grammar comment, otherwise LGTM.

[mojomex]

I've finally gotten around to test it! I used some gen1 taxi data and copied its concat parameters from one of our projects.

Looking good I think!
One question that might as well be me not understanding how to configure concat:
When deliberately messing up the params for one LiDAR, I would (intuitively) expect that there isn't a huge drop in concat output frequency, but rather that a group is dropped only sometimes, or every second time, etc.

I think this might be because we are always killing the oldest group, never giving it a chance to finish. It might also be because my timeouts are so long that almost all waiting collectors will be cleared because the inputs come at 10Hz.

Let me know your thoughts 🙇

Naive mode

small-compressed-naive.webm

Advanced mode, correct params

small-compressed-advanced_correct.webm

Advanced mode, incorrect params

small-compressed-advanced_incorrect.webm

[mojomex]

After talking offline, it is clear that a user can always misconfigure concat enough to look like the bottom video - that's the user's fault, but the algorithm itself seems to work fine 👍
LGTM