+1 That would be nice.
I want to know
1.how to localize robot indoor accurately except UWB
2.how to monitor the network normally avoid robot is out of online , and out of control
3.robot hardware is so boring , if there any simple way to construct a robot
Current ROS Navigation stack status summary:
- What’s the recommended basic configuration for a new robot nowadays.
- What alternative pieces we have (move_base_flex hello!) (and their purpose/advantages)?
- What alternative global planners are available?
- What alternative local planners are available?
- What alternative localization packages are available?
- What low level controllers are available (for different kind of robot bases)?
- Status for ROS 2.
Current ROS Manipulation stack status summary:
- What’s the recommended basic configuration for a new robot nowadays?
- What alternative packages we have for it?
- What planners are available (and what are they good for?)?
- How to move an arm in cartesian space?
- What whole body control stacks are available?
- How to deal with multiple arms?
- How to deal with exchanging end effectors?
- What pick-and-place pipelines are available?
How to prototype quickly some ROS package? (I could talk about how I do it if people are interested…)
- How to glue together ROS nodes quickly (I have a few tricks).
- Taking advantage of rqt tools, Rviz, dynamic reconfigure, params, launchfiles, rosbag…
- What does it take to convert it from a quick-and-dirty prototype to an actual ROS package expected to be used by all the community? (correct installation, roslaunch, examples, tests…).
- From my dirty script in my home folder to a released ROS package in multiple ros distros.
I’d be interested in seeing an MVP demo of a real application in ROS2 (more than basic pub/sub), and explaining some of the preconfigured options available or what’s coming in the near future.
Then explaining areas where help is requested so we can know what areas people are looking for outside contributors to help with specifically.
Many companies creating autonomous robots for unstructured environments are realizing that its impossible to deal with all the edge cases upfront, so they are opting to create remote operating centers where the robot asks for help, and a remote operator then remotes in to deal with the edge case. I’d be very interested on a talk on how this can be done with ROS.
How about of AI, anyone interested in hearing a bit more of how AI techniques are being used together with ROS and Gazebo?
I want to hear more about using Ros with deep learning and machine learning focussing on application such as navigation
Another suggestion: deployment approaches for ROS applications in commercial and research settings that are a little bigger than one or two robots. Containerisation, network management, runtime testing, internal buildfarms, quality assurance processes, telemetry, etc.
How about of art, many modern artists want to make autonomous system or robot to show what they thought. Talking about experiences co-working with them might be a interesting story and unique.
I’d like to hear an overview about testing: how to quick-start, what options are available, tips etc. - AFAI googled there has been no dedicated talk about testing in previous ROSCon occurrences. This kind of talk will help esp. new hires and people with less experience in testing in ROS and/or general development.
Some topics I can think of off the top of my skull. Some of these may be based on my misunderstanding / lack of understanding (I myself need the overview :))
- Integration of PyTest, newer testing framework in Python that is arguably? favored over
unittest, was recently announced by @machinekoder.
- How to start writing tests.
- Infra structure improvement discussion?
I would like to here more about ROS-Control concept in detail. Especially the basic concepts of it and how for example create standard control system from Sensor, Controller and Actor.
Currently are Sensors and Actors always part of the RobotHW which is in some cases very cumbersome to use. This is just a point I thought about a lot lately and did get to any sensible solution.
P.S. I am even eagar to research it myself until the presentation, but as of today my knowledge does not go deep enough.
This is perhaps a little last minute, but I’ve been back and forth between submitting a proposal for a few weeks now.
I’ve been working with testing in ROS, and have been tinkering with creating a new package to abstract unit and integration tests a bit further. In my mind, the package would allow you to simply define subs/pubs and test a given set of nodes to verify their output. Take for example a simple node that does filtering on command input data, you would provide example input to the node, and check for expected output. It would theoretically also be able to take rosbags as an input.
Would a talk briefing over the state of testing, and then going over this sort of package be valuable? I know I can get something working by then, but whether it’s a full on package or work in progress would to be seen.
I have been writing a lot of rostests and gtests lately and would be very interested in this project/proposal. Also please keep me informed about the project, I would love to contribute if it makes sense.
Good to hear, and will do! We fully intend to open source it early on and welcome contributions.
Continuing the testing and debugging theme, here’s the demo video in my proposed talk:
The talk is about using Mozilla rr for reversible debugging, coupled with some tricks to make debugging ROS nodes easier by republishing messages from inside the replayed process.
I love it, hoping to see your talk. This will hopefully help solve some Heisenbugs that are really hard to reproduce…
Thanks! The tool for republishing ROS messages is available at https://github.com/larics/gdb_ros_publisher.