Encountered a problem with TF conversion when using PointCloud Preprocess.

[Mr-MoNET]

Dear Autoware technical staff, I recently encountered some tricky issues while using the autoware_pointcloud_preprocessor package. Specifically, I encountered the following error when using the voxel_grid_downsample_filter node; I've been setting breakpoints and tracing the problem for a long time, but I still haven't been able to resolve it.
I also tried using AI tools for research, but I didn't find a good solution. My process was as follows:

1、ros2 launch rslidar_sdk start.py
Use this command to start the Robosense E1R, a solid-state LiDAR. This node will publish a topic called rslidar_points, which follows the Autoware LiDAR format with message format XYZIRCAEDT, type pointcloud2.

2、ros2 run tf2_ros static_transform_publisher 0 0 0 0 0 1 base_link rslidar
This command defines a static TF transformation because it's not currently installed on an actual device; the goal is simply to get the autoware core functionality package running, hence the use of a static TF transformation.

3、ros2 launch autoware_pointcloud_preprocessor voxel_grid_downsample_filter_node.launch.xml input_topic_name:=/rslidar_points
The following error occurred when I entered this command. Through breakpoint tracing, I determined that the problem was still caused by the TF conversion. However, I couldn't find more detailed information in the official documentation for autoware_pointcloud_preprocessor, so I don't know what to do next.

4、It's noteworthy that I didn't encounter any errors when running the approximate_downsample_filter node, and the first two steps were completed identically, also using static TensorFlow transformations, which is confusing me. I noticed two differences between approximate_downsample_filter.launch.xml and voxel_grid_downsample_filter_node.launch.xml. The former doesn't have default values ​​set for input_frame and output_frame, while the latter's default value is base_link. If I remove the default values ​​for input_frame and output_frame in voxel_grid_downsample_filter_node.launch.xml, the program stops throwing errors, but I still can't subscribe to valid PointCloud2 data on RViz.

zhsd@zhsd-B760-GAMING-X-GEN5:~/Desktop/autoware-humble$ ros2 launch autoware_pointcloud_preprocessor voxel_grid_downsample_filter_node.launch.xml input_topic_name:=/rslidar_points

[INFO] [launch]: All log files can be found below /home/zhsd/.ros/log/2025-12-16-16-34-54-428920-zhsd-B760-GAMING-X-GEN5-66520
[INFO] [launch]: Default logging verbosity is set to INFO
[INFO] [voxel_grid_downsample_filter_node-1]: process started with pid [66521]
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.593111421] [voxel_grid_downsample_filter_node]: start...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.595779669] [voxel_grid_downsample_filter_node]: Init tf...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.595893350] [voxel_grid_downsample_filter_node]: Init tf done...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.595940483] [voxel_grid_downsample_filter_node]: callback done...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.596016239] [voxel_grid_downsample_filter_node]: TF frame: base_link.
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.596065663] [voxel_grid_downsample_filter_node]: TF frame: base_link.
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.596079356] [voxel_grid_downsample_filter_node]: TF frame: base_link.
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.644618687] [voxel_grid_downsample_filter_node]: faster_input_indices_callback START...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.644796180] [voxel_grid_downsample_filter_node]: calculate_transform_matrix START...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.662840864] [voxel_grid_downsample_filter_node]: calculate_transform_matrix END...
[voxel_grid_downsample_filter_node-1] [WARN] [1765874094.721975874] [voxel_grid_downsample_filter_node]: convert_output_costly start...
[ERROR] [voxel_grid_downsample_filter_node-1]: process has died [pid 66521, exit code -11, cmd '/home/zhsd/Desktop/autoware-humble/install/autoware_pointcloud_preprocessor/lib/autoware_pointcloud_preprocessor/voxel_grid_downsample_filter_node --ros-args -r __node:=voxel_grid_downsample_filter_node --params-file /home/zhsd/Desktop/autoware-humble/install/autoware_pointcloud_preprocessor/share/autoware_pointcloud_preprocessor/config/voxel_grid_downsample_filter_node.param.yaml --params-file /tmp/launch_params_a91fn1ue --params-file /tmp/launch_params_0yx32lg4 -r input:=/rslidar_points -r output:=/sensing/lidar/top/voxel_grid_downsample_filter/pointcloud'].

[liuXinGangChina]

我用的也是 速腾雷达，没遇到过你这个情况，录个bag发给我看看，2秒钟就可以

[Mr-MoNET]

error_rog.zip
我使用的是robosense的E1R型号，全固态激光；autoware官方好像用的是机械式那种，我录制了这些话题。
/rslidar_points：lidar输出原始点云数据，pointcloud2类型，我们改了驱动，字段不是默认的XYZI，而是适配autoware的XYZIRCAEDT
/tf_static：我们使用了一个静态的TF转换，命令是ros2 run tf2_ros static_transform_publisher 0 0 0 0 0 0 base_link rslidar
//sensing/lidar/top/voxel_grid_downsample_filter/pointcloud：autoware点云下采样的话题，但是这个话题运行的时候会报错，详细情况是；如果我不启动/tf_static，他就会报base_link转换的错误，如果我一旦启动了静态转换，他就报错了，报错信息是：

[Mr-MoNET]

麻烦大佬帮忙分析一下原因，真是没招了；早上我把yaml配置文件的input_frame和output_frame设置成这样，然后不报错了；但是rviz收到数据后，无法正常显示点云：

就是说我在启动静态TF转换后，我把转换的base_link作为input，因为autoware会对input_frame做转换，转换后对应的输出就是output_frame。但是rviz这里只有提示数据类型不匹配。所以我在想有两种可能：
1、因为改了lidar驱动，输出数据类型还是有问题造成对不上
2、要么就还是静态TF转换的问题，但是如果静态TF转换有问题，为什么我修改yaml后就不报错了，这个说不通

[liuXinGangChina]

ros2 launch autoware_downsample_filters voxel_grid_downsample_filter_node.launch.xml input_topic_name:=/rslidar_points output_topic_name:=/downsampled_points input_frame:=rslidar output_frame:=base_link

我看了你的bag，没问题的。你用这个指令
最近比较忙，社区消息可能看不到，实在搞不定加我微信吧

黄色是下采样以后的输出