Closed osrf-migration closed 7 years ago
Original comment by Nathan Mertins (Bitbucket: nmertins).
We have identified this issue and are in the process of integrating a fix. This problem should be resolved in the next release of the IHMC controller.
Original comment by Jordan Lack (Bitbucket: jordanlack).
I'm pushing out a release today that should make the hands function. Keep in mind the way the SRCSim package is packaged up before you get it though it may take some time for the OSRF folks to take what we give them and put it into an updated release of their own.
Original comment by Erica K (Bitbucket: Erica K).
Having spent the better part of a day trying to get a hand parallel to the wall, I would like to check on the status of this. Has it been released, and if so, how do I get it?
Original comment by Jordan Lack (Bitbucket: jordanlack).
The NASA code has been pushed that exposes the ROS control interfaces to controllers for controlling the fingers. Since we are not able to make our debians publicly available ourselves, it wouldn't be available unless the OSRF folks have grabbed our updated debs and put them into a release of the SRC sim. I am not sure if that has happened yet or not. @nkoenig or @scpeters do you know?
Original comment by Erica K (Bitbucket: Erica K).
Thank you for asking if the finger code has been pushed. I would also like to know if the forearm and wrist joints are now able to move via an ArmTrajectoryRosCode and HandTrajectoryRosCode, as that is also highly relevant for us.
Original comment by Nathan Mertins (Bitbucket: nmertins).
We have not yet made a release containing the fix for enabling forearm and neck control. Currently our plan is to have that out tomorrow or Wednesday.
Original comment by Rud Merriam (Bitbucket: rmerriam).
Wait a minute. In Issue #60 it was stated only bug releases would be released. This is not a bug release but a new capability despite the importance. You're now going to be penalizing those who spent the time getting qual2 to work without this capability. Now we're going to have to revisit this part of qual2 instead of working on improvements in other aspects.
Original comment by Erica K (Bitbucket: Erica K).
It is hardly a new capability. It was advertised in the API, and was even in the arm demo in the setup tutorial -- but the lower limbs were silently failing all the while. We wasted a huge amount of time on this.
Also, note the phrase "containing the fix".
Original comment by Rud Merriam (Bitbucket: rmerriam).
You're playing semantic games, Erica. At this point it's a new capability.
Introducing this level 'bug' fix less than two weeks before the deadline is not appropriate. In Issue #60 I specifically asked when we would see a freeze so we could plan on how to proceed.
I just noticed this include neck control which means qual1 is affected, also.
Original comment by Erica K (Bitbucket: Erica K).
I strongly disagree. This capability was in the documentation from the beginning. Not to mention that I believe the purpose of the competition is to show our effectiveness in using NASA's controller, not writing our own. A robot without working hands, wrists, or forearms, is hugely crippled.
In any case an argument between us seems irrelevant, as the bug fix is in progress it seems.
Original comment by Erica Tiberia (Bitbucket: T_AL).
Though the capabilities of neck/ hand/ wrist etc were advertised, they were unavailable for the majority of the active qual completion time. A fix for this at this stage of the game, where some teams have already submitted log files, I do think constitutes a new capability.
I agree that this problem has been extremely crippling, but at least it was an even playing field for everyone. This is penalizing to people who front loaded their competition work.
Original comment by Rud Merriam (Bitbucket: rmerriam).
This discussion is important because those of us who moved ahead and got everything working without this capability are going to be penalized. This discussion is making all the viewpoints visible to the competition.
I didn't develop a new controller because (1) there aren't enough hooks to do so and (2) the competition specifically said you can't. I worked with what I had. I put off the button press work on qual2 waiting until the Issue #60 comment. Then I made it work with what existed.
Original comment by Erica K (Bitbucket: Erica K).
I doubt there are even 20 teams actively competing at this point. I am also surprised at the cut-throat nature expressed here. "Visible to the competition?" What, we should let people waste days of their time working on their code when the API is the problem? For what it's worth, our team can get the robot through without a working API also, and may well have to leave it that way given the slowness of these bug fixes -- but it's ridiculous. No one would ever use a robot this way.
The code is broken. When it is fixed all of us are getting the same code. I doubt either of you have submitted your log files, of course you will keep tweaking to get the best possible run time and submit as late as possible.
Original comment by Rud Merriam (Bitbucket: rmerriam).
I submitted a qual2 log a week or so ago but am still tweaking so may replace it. For qual1 I have a good log I could upload but want to try a couple things first.
No, it's not cut throat. I've been helping in general ways anyone who asked questions here or in the Facebook group.
It's unfortunate but these competition tend not be be about doing things the right way but about winning the competition. I don't think NASA is going to get the best teams for the actual competition out of this qualifying run. I include myself in that category should I qualify. I'm good but not the best because I have gaps in my knowledge about ROS, OpenCV, and other aspects of robotics. I'll be fully head down working on those gaps until disqualified.
Original comment by Nathan Mertins (Bitbucket: nmertins).
Speaking on behalf of IHMC, I understand your frustrations regarding the lack of some core functionality, namely control of the neck and forearms. We will examine our (IHMC's) internal release pipeline and come up with a better way to keep you all informed of progress towards SRC releases and what those releases will contain.
Original comment by Paul Ferretti (Bitbucket: paulf1960).
Seems like a bug fix to me. The API mentions it from the beginning. It's not like we suggested that they create this functionality just for our team. I think a working API should have been provided from the beginning but I'll settle for getting one now.
Original comment by Richard Isakson (Bitbucket: RichardIsakson).
I tested the neck and lower arms today. They worked well.
Issue resolved.
Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).
Original report (archived issue) by Richard Isakson (Bitbucket: RichardIsakson).
I have spent a bit of time testing the robot's arms using both the files arm_traj_yaml.py/Traj_data.yaml and the file armDemo.py. While the shoulder joints, upper arms and elbow joints work quite well, the lower arms and wrists don't work at all. I haven't tested them, but I'm told the hands don't work either. I also tested the neck using the files neck_traj_yaml.py/Traj_data.yaml and it still doesn't work. While it may be possible to complete the Qual2 trail by bludgeoning the door switch with the dead stumps, if the switch worked, it won't be possible to continue the challenge to the more advanced level with the arms in their current state.
The machine is a quad core I7 with 16 gigs of RAM.