Feature/refactor covariance displays #11
Conversation
… into position and orientation
Now it's possibile to fix the visualization on the pose
This class will exhibit the covariance properly given a covariance 6d covariance matrix (float64 covariance[36]) from a message
While here, add visibility parameter to constructor
…iances Since we use the scale to set the covariance shapes, we had to add geometry properties for the arrow like it is done with rviz::PoseDisplay (head and shaft properties separated), instead of just a scale value like it's done in rviz::OdometryDisplay.
Also, while here fix includes of rviz files. Rationale behind this commit: The arrow class like it's defined points to -Z, so we were rotating it to make it point to +X. The problem is that if we make the arrow's node parent of covariance's nodes we'll have the covariance rotated as well. The work-around is to keep a deque of scene nodes, and make them parents of respective arrow and covariance nodes. With this arrow's node can be rotated without affecting the rotation of covariance's nodes.
…ild_frame_id through TF
Ellon
force-pushed
the
feature/refactor_covariance_displays
branch
from
db9d630 to
42faf6a
Compare
This allow us to draw the position covariance wrt in the frame defined by message's frame_id, while the orientation covariance is still drawn in the "child frame". The frame node should be positioned externally (by the displays holding covariance visual objects).
|
Looks good to me |
The scale should be twice the standard deviation. This way the scale 1.0 will represent one standard deviation. While here, fix the parameter name (shape -> scale).
|
Can this be merged and then any subsequent improvements be proposed in new PRs? Integrating this would allow this package to be used, which is not possible ATM due to the wrong results of the ellipse orientation. |
|
@v01d It doesn't look this thread is being watched by any maintainers... Maybe @thomas-moulard or @mherrb could take a look at it? |
|
If Thomas is still interested, let him validate this. If he doesn't answer in a few days, I will take care of it (or just give you access to the repository) |
|
Hi all, thank for waiting for my input! @mherrb could please go ahead and merge this? I don't have any write-access to the repo anymore and I don't use ROS for my work currently so I don't have the bandwidth to maintain this plugin. IMHO the best thing to do is to send this as a PR to rviz and see if we can get this included by default. Last time we checked the maintainers seemed to be willing to accept it. |
|
BTW if rviz integration is not an option but someone is interested to maintain this code. Please check with @mherrb . On my side, I don't have any objection to have anyone take over this repo. Thanks! |
|
Hello Thomas, thanks for the input. I plan to send this as a PR to rviz later, after I have the angular covariances being displayed properly. |
|
Hi guys. I found my way over here from ros-visualization/rviz#974. My slight preference is to get this working here first and then propose merging this into rviz after people agree that it is working correctly. The idea being that people who are already using this as an external plugin can benefit from the improvements too. I'd probably only consider this plugin for merging into Jade or Kinetic. If merging went well for those versions then I'd consider Indigo, but it won't be a quick turn around. Thanks! |
|
(Sorry I miss-clicked to close the PR) OK @wjwwood I'll update ros-visualization/rviz#974 when this PR is merged. I'm not sure if there's anyone using this plugin apart from myself and maybe @v01d . I think the agreement will probably come from two of us. :) |
Feature/refactor covariance displays
|
Hi Ellon, I've merge the request and added you as team member for this repository. You can now manage it directly. |
|
Ok @mherrb , Thanks! |
|
Looks like I'm late to the party... @Ellon: We are using this plugin. Thanks for the update, i will give it a try shortly. |
|
P.S.: +1 for getting this into rviz proper, even indigo, and discutinuing development here. |
|
Now that the issues are sorted out here, I'll take a look at the rviz pr. I just wanted to separate the concern of fixing the thing and getting it put into rviz. Also, since I take on some level of maintenance when things are merged into rviz, so I tend to review it carefully before rolling it in. |
|
Er... in fact there are still some issues I just opened, mostly related with 2D poses and rotational covariance. Please check #10 and the currently open issues. |
In this PR I re-factored the plugin's Pose and Odometry displays to enable back all the functionalities from the original rviz displays, in a way that the covariance looks like more or less like an addon to the original displays. This PR is based on the code from #9.
The computation of the covariances are the same as before, but just wrapped into one file (covariance_visual.cpp). The covariance visualization is also the same (ellipse for position and a cone (?) for orientation). I think this way it's easy to improve the drawing of the covariances later.
One thing I did was to add some functionality from Pose to Odometry, like possibility to visualize the odometry as arrow or axes, and a finer control of the shapes being draw (axis length and radius, and arrows shaft/head length/radius). The selection tool for the pose with covariance is also working.