fermi
fermi copied to clipboard
ros-industrial-consortium
Better integration into MoveIt repos?
This appears to be somewhat robot-specific at the momement, with all the fermi prefixes, etc. Are there plans to move this into the main MoveIt! repos and have it part of the full release?
@davetcoleman most definitely there is a plan for this. I have chatted with @shaun-edwards about this before and I think it would be very nice to have this plugin as part of the main MoveIt! release. Currently I am focused at improving and bug fixing some of the components of the plugin and also working on the indigo version on it. There are some issues which I hope I can solve them before proposing to move this plugin into the official MoveIt! repo. @shaun-edwards any opinions regarding this?
Cool, let me know how I can help
Also, I was thinking about testing this today, but browsing quickly through your documentation it was not clear to me how to use this for an arbitrary, moveit-configured robot
@rkojcev I'm all for moving this functionality into MoveIt, but clearly there are improvements to be made. We have been able to show the plugin working on different robots, but there appears to be an implicit requirement that the robot group be named manipulator. Since all ROS-I robots follow this convention, we didn't notice. This among other issues are in the plans to address.
@davetcoleman Your help would be much appreciated, since you probably have a clearer idea of how make this tool truly robot agnostic. Perhaps you could point us to a non-ROS-I robot, that we could try to exercise.
thanks a lot @davetcoleman for your comments and suggestion which robot we might test.
Maybe you could look at these tutorials. If you have any problems with getting up and running this plugin please let me know. I agree with @shaun-edwards that one of the things that I would like to change and improve is this dependency of the hard coded "manipulator" robot group name.To be honest I have kinda forgot about this one and I have just got reminded today by a question posted on theROS-I mailing list. Also the Cartesian path is better executed when you are using the ROS-I simulator. This plugin was only tested on Industrial manipulators and I have not tested it on other type of robots like Baxter and PR2 for example but it would be nice to see how the plugin behaves with these types of robots.
These are the couple of the things I have in mind to improve, as I move along making these improvements, better tutorials and documentation, etc. when I will try to find an issue I will address it on this repo. I would be really happy if I get further more suggestions for improvements and bug fixes from you guys as well :)
Oh I didn't see that - could you put the tutorial link on the github readme?
Done!! you can find the link to the tutorial at the end of this readme
Thanks a lot for your comments, feedback and help.
I managed to get it work with Denso robot https://github.com/start-jsk/denso/pull/48 The concept loos really nice (although I had to spend some time on playing with its user interface, and getting this to work with the new robot)!
Hi @130s , I am very happy that you managed to get this plugin working with the Denso robot. I have made new pull request where few issues have been resolved and some improved. I haven't manage to get into @shaun-edwards review yet ( due to lack of time at this moment). Plus there should be some update of the tutorial + documentation.
Please bare with me there will be new update of this repo soon :)
Hi @rkojcev, I was surprised to see this week that several groups in Singapore were using Fermi to create Cartesian waypoints in MoveIt! - it was pretty cool! I would still like to see this better integrated as a main MoveIt! feature - are you still active with ROS / this repo?
Hi @davetcoleman
That is very good news. I am really happy that people are using this repo. I have made few updates on this repo here. The most active development I am working on now is integrating my current GSoC project into fermi in this branch. This extends the current plugin to allow the user to set Cartesian Impedance and Force Control parameters based on the new message set I am working on.
I would be happy to make the effort towards better integration in moveit. If you have time it would be great if you have a look and let me know what you think.
Thoughts on majorana msgs:
- Is it possible to have fewer message? e.g. reuse existing messages instead of creating multiple custom ones for a i.e. 2x3 matrix (CartesianDamping) or a message that just wraps a Twist (MaxCartesianVelocity)?
- Are you planning on polling the greater ROS community to find people with experience with impedance control to get their feedback? I haven't worked on this recently, or much, so do not have the best feedback for standard message types
Thoughts on fermi:
- Would probably be easiest to add to a unified moveit repo: https://github.com/davetcoleman/moveit Hopefully this happens soon
- Would be best to demonstrate that it is agnostic by simply having it run on two different robots
- Its easiest to review code when it is in the form of some sort of pull request
Thanks a lot @davetcoleman for the info. I have moved your comments regarding the impedance msgs into the pull request for the same repo. I will comment there and probably it will be better to have a discussion there as well.
regarding fermi: I will take a look at your suggestion and probably clean up fermi a little bit. There are lot of examples which moveit packages are depreciated, #34. @shaun-edwards is this ok with you?
I can confirm that this plugin works on kuka iiwa right out of the box.
Anyone else who tested this on a real robot can confirm or if had any problems setting up can open issues on this repo.
Hit the button too early :/
I believe there's a great need for this GUI-based Cartesian planning!
From my 2-year old experience https://github.com/ros-industrial-consortium/fermi/issues/26#issuecomment-76895034 this package can be used as a library and no source code development isn't necessary. So making a binary release IMO will be absolutely beneficial for potentially many users.
Since I'm very interested in using fermi for my robots, I can help if there's anything I can do. Since Dave and I are maintaining MoveIt! these days, incorporating into there shouldn't be a problem if we decide to do so.
Hi @130s unfortunately I do not have much time working on this at the moment. As far as I know there is some modifications that need to be made for the plugin to work on the ROS Kinetic version.
If someone could modify the plugin to be compatible with the Kinetic version I would welcome pull request to this repo.
As far as I know there is some modifications that need to be made for the plugin to work on the ROS Kinetic version.
How about starting the migration with ROS Indigo? For Kinetic, once the necessary work we can migrate too.
This issue is labeled as a "Simple Improvement" under issues in the moveit repo, so hopefully someone will try to tackle it next week for World MoveIt! Day. Perhaps you? :-)
@davetcoleman thanks to @ahcorde, and to https://github.com/ros-industrial-consortium/fermi/pull/48 the plugin has been migrated to melodic.
Whoever is interested can review it and test it out.