Here are the notes I took, please correct anything I got wrong/
2020-10-22 Maintainer meeting
Jeremy Adams - St Luis, Object Computing
Works on thier own DDS implementation that isn’t released publicly yet. On working group. Trying to position themselves to help with DDS middleware specification and implementation.
@AndyZ could you respond with an update on this?
Mobile manipulator. For ROS2 Galactic there are a number of platforms that were selected for demonstration. One is a 4 legged walking robot that doesn’t do manipulation and Hello Robot that does manipulation. Interesting kinematics and we (picknik) has entered in a partnership to help them make the default learning platform for ROS2 mobile manipulation. Think of it as the turtle bot of mobile manipulation. Not a PR2, not an industrial robot, but is aimed at a different market and should be a good package to play around with ROS2. Does anyone else have any other hardware platforms they are working on?
UB1, not going to be good for the general public because there is only one of them.
Anything you do for the UB1 is probably generally useful.
Any work on hardware
Noetic and Melodic release
Noetic was realased and I am planning on working on Melodic release. Any help would be greatly apreciated.
We could do better at making it easy to get started using the library.
It all sounds good, what extra work are we are taking on if we are doing this. This would be continuing work.
I think it’d be hard to identify issues like this.
Are people feeling positive about this? We can create the landing page for this and start lableing issues. Assuming we can get the technical issues figured out what are people’s thoughts?
Would it be useful if I made a PR to moveit.ros.org?
Looking at website we are missing details on the pull request page to guide people through becoming first time contributors.
What are the action items?
Tyler will make a draft webpage and we’ll embed issues.
I’ll do that and create an issue on moveit to track that.
(Notes from Felix)
Stuff I think we should do better in terms of setting up the environment:
- make the environment variables like CATKIN_WS work in the master-source docker container so clang-tidy commands can be copy-pasted directly from the instructions
- add convenience scripts like bash aliases to run clang-format/clang-tidy in the docker container
- add simple instructions to use docker, including a way to link a workspace that’s visible on the host hard drive (many first-timers are not fluent in linux+CLI+docker)
Bloom issue is fixed so we are ready to bloom-release for foxy. All the dependencies are synced up. I still want to do another sync for MoveIt main branch. Nothing is blocking it and all the demos are still functional. We are
New features are warehouse is now ported, you can now use that in rviz.
We have a new intern, Sebastian working on a hybrid planner. That is a combination of a global planner that will plan around objects quickly in a course way and a local planner that can iterates between points. We are just starting on research on what sort of applications we want serve with it. We are also in the early stages of architecture for it and next meeting we’ll invite Sebastian to give and update. The first prototype might be the global planner running Trajopt or Stomp or RRTConnect and the local planner be Jacobian. We are still working on architecture to make it as flexible as possible. He will start an issue and present is research at the next time. We’d like to see people communicate to him about what sort of applications it should work for.
Does he have a specific demo in mind?
We don’t have that yet but the first demo should probably be collision avoidance with a changing scene. Global planner has to replan and local planner adapts and follows new plan.
What does Toyota care about moveit?
We are using Drake. We are interested in ROS2. I am also interested in this because I used to be a MoveIt maintainer.
Since you’ve spent time in both Drake and MoveIt community do you have feedback for MoveIt.
I think MoveIt’s strong points are how it integrates with ROS environments. I think it also really good at first time use with new robots. Drake has a high learning curve and we would like to improve there.
My understanding is Drake is coming from control and dynamics.
I think that is true and there is really cool advanced optimization and simulation tools that tackle those problems from a different way. There are also math tools that let you do symbolic representations that you can use to do advanced control systems.
When I looked at drake I saw the math stuff really the strong suit.
Drake tries to do everything, for better or worse. That can be useful if you are doing everything inside Drake, but if you are trying to connect other things it can be hard. There are three core parts to drake, motion planning, math modeling, and simulation. There is some cool stuff in there. I am more familiar with MoveIt but I am now coming up to speed with Drake as it part of my job.
Is there interest in growing the open source community for Drake.
There is interest but not like what you are doing for MoveIt. I plan on coming to this meeting when I can fit in my schedule.
Please speak up if you see issues where we can learn from each other (MoveIt and Drake).
Related to Drake, I think one of the coolest features we could do for MoveIt would be symbolic math features.
Redesign of Get involved page
Shared google doc with redesign of “Get Involved” page.
Love the buttons.
If you look on our website you will see the roadmap it is basically the Rosin grant. I would like to get the maintainers to contribute what their ideas are for the next couple years so we can publish a roadmap.
Last we did an exercise at MoveItCon and we should recreate that virtually.