pointcloud to laserscan slow

asked 2016-10-26 07:14:09 -0500

gijsje170 gravatar image

updated 2016-10-26 07:28:24 -0500

gvdhoorn gravatar image


For my project i am using a raspberry Pi which is connected to a kinect. I wan to use the data of this kinect for driving around the building. The only problem with it is that the pointcloud_to_laserscan node only gives out a laserscan once every 15 seconds. This is way to low. Does anyone how to easily fix this problem. Using Raspberry Pi 3 ROS kinetic Kinect v1

edit retag flag offensive close merge delete



If you're building pointcloud_to_laserscan from sources, be sure to enable optimisations (-DCMAKE_BUILD_TYPE=Release). If not: get a platform with some more processing power? Pi's are quite limited, and if you're not using nodelets, the pointcloud gets copied quite a nr of times.

gvdhoorn gravatar image gvdhoorn  ( 2016-10-26 07:29:29 -0500 )edit

Also: note the comment on the p2l wiki page:

If you're trying to create a virtual laserscan from your RGBD device, and your sensor is forward-facing, you'll find depthimage_to_laserscan will be much more [..] efficient since it operates on image data instead of bulky pointclouds.

gvdhoorn gravatar image gvdhoorn  ( 2016-10-26 07:31:03 -0500 )edit

Thanks for your help when building with -DCMAKE_BUILD_TYPE=Release it improved my performance. Currently it is giving me an average of 4FPS which is quite nice already.

gijsje170 gravatar image gijsje170  ( 2016-10-26 07:44:01 -0500 )edit