feat(autoware_lidar_frnet): integrate cuda blackboard into point clooud process

[kyoichi-sugahara]

Description

This PR integrates CUDA blackboard into the autoware_lidar_frnet package to optimize point cloud processing performance.

Main changes:

• Modified to use CUDA blackboard for direct processing of point cloud data on GPU memory
• Changed memory copy for input data:
  • Before: cudaMemcpyAsync(..., cudaMemcpyHostToDevice) - CPU→GPU transfer
  • After: cudaMemcpyAsync(..., cudaMemcpyDeviceToDevice, stream_) - GPU→GPU transfer
    This eliminates CPU↔GPU memory transfers, reducing latency and CPU load

This reduces latency and CPU load in the point cloud processing pipeline, improving efficiency especially in workflows centered around GPU processing.

Related links

Private Links:

• TIER IV

How was this PR tested?

• Ran the entire sensing module with /sensing/lidar/concatenated/pointcloud as input, operating in the same container
• Played back actual rosbag data and verified that the "negotiation message has been received" log message was successfully obtained as follows:

Notes for reviewers

• The input topic is subscribed via CUDA blackboard, so upstream nodes must also support CUDA blackboard
• Important: The subscriber node must be in the same container as the publisher node. This is a prerequisite for CUDA blackboard to work correctly, as it relies on shared GPU memory within the same process/container
• The output topic message type remains unchanged (sensor_msgs::msg::PointCloud2)
• If the input point cloud size is large, you need to modify ml_package_frnet.param.yaml and regenerate the engine

Interface changes

Topic changes

Modifications

Version	Topic Type	Topic Name	Message Type	Description
Old	Sub	~/input/pointcloud	sensor_msgs::msg::PointCloud2	Input pointcloud (CPU memory)
New	Sub	~/input/pointcloud/cuda	sensor_msgs::msg::PointCloud2 (via CUDA blackboard)	Input pointcloud (GPU memory via CUDA blackboard)

Note: The topic name for negotiation is now ~/input/pointcloud/cuda (with /cuda suffix added by CUDA blackboard), but the user-facing topic name remains ~/input/pointcloud. The message type remains unchanged, but the internal implementation now uses CUDA blackboard.

ROS Parameter Changes

None.

Effects on system behavior

• Performance improvement: Reduced CPU↔GPU memory transfers, improving processing latency
• CPU load reduction: Eliminated need for copying to host memory, reducing CPU load
• Compatibility: Output topic interface remains unchanged, so there is no impact on downstream nodes

[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

❌ Patch coverage is 0% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 17.83%. Comparing base (9d55964) to head (9856c27).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...tion/autoware_lidar_frnet/src/lidar_frnet_node.cpp	0.00%	6 Missing ⚠️
...r_frnet/include/autoware/lidar_frnet/ros_utils.hpp	0.00%	3 Missing ⚠️
...erception/autoware_lidar_frnet/lib/lidar_frnet.cpp	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #11677      +/-   ##
==========================================
- Coverage   17.85%   17.83%   -0.02%     
==========================================
  Files        1746     1747       +1     
  Lines      122187   122319     +132     
  Branches    42890    42953      +63     
==========================================
  Hits        21818    21818              
- Misses      82086    82218     +132     
  Partials    18283    18283              

Flag	Coverage Δ		*Carryforward flag
daily	19.91% <ø> (ø)		Carriedforward from 5a5aec9
daily-cuda	17.80% <ø> (+<0.01%)	⬆️	Carriedforward from 5a5aec9
differential-cuda	0.00% <0.00%> (?)
total-cuda	17.83% <ø> (+<0.01%)	⬆️	Carriedforward from 5a5aec9

*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.