feat: gpu ground segmentation

[anhnv3991]

Description

@badai-nguyen created an interesting PR that uses CUDA to accelerate the ground-segmentation but the code is not quite optimized and well-organized. I re-organized his code a little bit and fix some minor bugs. Besides, I added some utility functions for allocating/freeing GPU memory.

Related links

Parent Issue:

• Dai's draft PR

How was this PR tested?

• Same as Dai's PR.
• Execute the cuda_scan_ground_segmenation_node for example:
  ros2 launch autoware_ground_segmentation_cuda cuda_scan_ground_filter.launch.py input/pointcloud:=/sensing/lidar/concatenated/pointcloud
• Open Rviz2, add the topics /cuda_scan_ground_segmentation_filter/output/ground_pointcloud and /perception/obstacle_segmentation/pointcloud. Those publish the ground and non-ground points on rviz. Set the global options Fixed Frame to base_link. Set the Reliability Policy of Point Cloud topics to Best Effort.
• Play rosbag which include /sensing/lidar/concatenated/pointcloud topic
  A sample t4_dataset whose input_bag with /sensing/lidar/concatenated/pointcloud topic available could be downloaded from here
• Download link

Notes for reviewers

None.

Interface changes

None.

Effects on system behavior

None.

[badai-nguyen]

Note:
I add 35c6835 to add option testing this node together with perception pipeline by enable use_cuda_ground_segmentation

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 19.40%. Comparing base (97333b0) to head (b3561cb).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #11371      +/-   ##
==========================================
+ Coverage   18.88%   19.40%   +0.51%     
==========================================
  Files        1680     1689       +9     
  Lines      114844   115612     +768     
  Branches    39690    40038     +348     
==========================================
+ Hits        21694    22435     +741     
- Misses      75015    75562     +547     
+ Partials    18135    17615     -520     

Flag	Coverage Δ		*Carryforward flag
daily	21.34% <ø> (+0.01%)	⬆️	Carriedforward from 9dfd826
daily-cuda	18.89% <ø> (+<0.01%)	⬆️	Carriedforward from 9dfd826
differential	5.76% <ø> (?)
total-cuda	18.89% <ø> (ø)		Carriedforward from 9dfd826

*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.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Documentation URL: https://autowarefoundation.github.io/autoware_universe/pr-11371/
Modified URLs:

• https://autowarefoundation.github.io/autoware_universe/pr-11371/perception/autoware_ground_segmentation_cuda/