Closed zhanghua7099 closed 2 years ago
By the way, can you provide some details about running KITTI dataset like your demo videos?
Hi!@zhanghua7099 Thanks for following our work. Is there any output in the GEM terminal? Can you provide a screenshot of the terminal? I'll try to find out what happened.
I try to debug it through rostopic echo /elevation_mapping_0/robot0/visual_map
. This topic didn't publish any messages. The similar situation occurs at these topics:
After I run the command:
roslaunch elevation_mapping_demos simple_demo.launch
This may be the reason why the elevation map is not displayed.
The point cloud of the lidar sensor can be visualized in rviz.
@zhanghua7099 You can return to the commit a359d011bf5f11e26cf0e330ecc4a3e82f590a76. I'll try to fix this bug ASAP.
I have tried the commit a359d01. It still has similar errors.
Thank you for helping me!
In the commit a359d01, the error message will be printed after I play the rosbag.
In the current commit, the error message will be printed after I run the command directly:
roslaunch elevation_mapping_demos simple_demo.launch
Hi @zhanghua7099 I test the newest version on my own computer and everything goes well. There are some changes in ROS noetic since I just test it on Kinetic and Melodic. Or you may try to update the repo since I've added some code to fit c++14.
Hi!
Thank you for your hard work. I have tried the new version. After I change the launch file, the code can be run.
Unfortunately, the updated version still has some errors related to CUDA programming:
addKernel launch failed: an illegal memory access was encountered
I didn't know what's wrong with my computer. I use CUDA 11.3 with cudnn 8.2.
@zhanghua7099 You can try to reboot the machine. If it doesn't work, you may have to change the parameters in filter_kitti.launch to downsample the point cloud to < 20000 points.
Hi! I changed the compilation options of CUDA. This CUDA error could be solved.
Just change the arch and code of the GPU can fix this error:
SET(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-O3;-std=c++14 -gencode arch=compute_86,code=sm_86)
The index of GPU could be checked on the nvidia website. It will be helpful for using RTX 30XX series.
By the way, I recommend adding C++14 to all CmakeLists files if using PCL.
Thanks a lot!
Hi! I changed the compilation options of CUDA. This CUDA error could be solved.
Just change the arch and code of the GPU can fix this error:
SET(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-O3;-std=c++14 -gencode arch=compute_86,code=sm_86)
The index of GPU could be checked on the nvidia website. It will be helpful for using RTX 30XX series.By the way, I recommend adding C++14 to all CmakeLists files if using PCL.
Thanks a lot!
I got the same problem, and I tried to change the arch and code of the GPU but catkin make failed.
Can you help me?
Thanks a lot.
@leslieburke Hi, can you paste the complete error information, it may appear at the beginning of the build process.
Hi! Thank you for your good work.
I built this project, but the visualization may have some problems for my computer. I use your provided
test.bag
, but I only get odometry results. The point clouds didn't have any visualization. The rviz received the pcd message, but it can't output it.Could you please tell me how to fix this bug? I use Ubuntu 20.04 with ROS noetic.