Efficiency improvements and customizable occupancy 2D map projection. #9
Conversation
+ Efficiency improvements: | avoid transforming clouds already in map frame (added additional parameter to specify sensor frame_id for calculating sensor pose) | create occupancy cloud directly as ros message instead of pcl cloud (avoids conversion) + Projection of nav_msgs::OccupancyGrid with customizable height and initialization as free or unknown. + Increased tf listener buffer to 30 sec.
|
Thanks for your contribution! Is this meant for hydro or indigo? Best send the pull request against the right ...-devel branch then in the future (master tracks the latest stable releases). More comments in the code. |
| size_t numberReservedPoints = m_octree->size(); | ||
| float* pointcloudDataPosition = NULL; | ||
| if (publishPointCloud) { | ||
| cloud->header.seq = m_numberCloudsPublished++; |
There was a problem hiding this comment.
This line (and the variable m_numberCloudsPublished) is not needed, the publisher should automatically increase .seq in the header.
of sensor frame id. - Line 310 of OctomapServer.cpp was doing an unnecessary internal copy (pc_nonground = pc), that can be avoided by using shared pointers. - Improved assembly of pointcloud from octree by using vector reserve instead of resize. - Improved parametrization of sensor frame id. This way the input cloud can be sent in any frame id (map, sensor, other), and the parameter sensor_frame_id can be used to specify which is the sensor frame to compute the origin for the raytracing algorithms (used to integrate the pointcloud into the octree).
|
Latest commit addressed the suggestions made during the pull request discussion. P.S. |
and regulate how often the 2D projection map is published.
|
Can I ask why this PR hasn't been merged? |
|
To be true, I haven't had a chance to test it thoroughly since the changes are rather fundamental with potential to change / break existing behavior. There are also multiple features & improvements mixed in one PR, which makes it hard to track the individual influence (as opposed to quickly merging a more atomic PR). |
|
I see and I completely understand :) I was just curious, as efficiency improvements are always wanted. |
|- avoid transforming clouds already in map frame (added additional
parameter to specify sensor frame_id for calculating sensor pose)
|- create occupancy cloud directly as ros message instead of pcl cloud
(avoids conversion)
initialization as free or unknown.