Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Re-create iris+quadrotor mi-convex planning examples #6243
referenced this issue
Jun 2, 2017
The reason for removing MATLAB Quadrotor code is that Drake's older mechanisms for MATLAB support are being removed. In particular,
I can hold off on the Quadrotor removal as long as possible if that helps, but that may only buy us 1-2 weeks, depending on how fast other turn-down PRs are moving.
In terms of keeping paper citations useful, I am open to suggestions. The easy starting point would be to add a
(Responding since I am passionate about "Reproducible Research" <https://en.wikipedia.org/wiki/Reproducibility#Reproducible_research>) For the future, I suggest providing a SHA along with the repository in the citation.…
-------------------------------- Evan Drumwright Senior Research Scientist http://positronicslab.github.io Toyota Research Institute Palo Alto, CA
On Tue, Jun 6, 2017 at 9:06 AM, Jamie Snape ***@***.***> wrote: FWIW IRIS and its Python bindings build with Bazel easily enough: rdeits/iris-distro#57 <rdeits/iris-distro#57> if Python were a suitable alternative to MATLAB. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#6243 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACHwz5Fl_6rUwA8Al6G-vCD94nm5ZXIzks5sBXjzgaJpZM4NuxOK> .
1 or 2 weeks doesn't matter either way, so if RigidBodyManipulator is going then I guess the Quadrotor code is going too. Having a readme with a link to the relevant branch would be nice.
@jamiesnape I would love to have the example in Python rather than Matlab. But converting it would be equivalent to rewriting it from scratch, since the APIs are all different
@edrumwri yeah, lesson learned.
But I have to say that I'm still not happy with this situation. My understanding was that the high-level APIs in Matlab would remain, but be powered by the new C++ core. Deleting RigidBodyManipulator without any replacement means that a great deal of Drake's scientific content will cease to work, and it risks alienating Drake's long-time users.
@rdeits Someone whose opinion I value very much convinced me about two years ago that the code we'd been developing was formatted/packaged in a way that made it inaccessible to many and essentially unusable by anyone working with real robots. I agree it serves as a runnable example for those reading and understanding the papers, and that remains true (on the old sha).
I'm personally committed to taking the scientific content developed here to maturity in the new framework because I believe that it can and should be used by many. (I asked Jeremy to create this issue and assign it to me). I think that the new code is far enough along now that we're ready to zap and port.. and not zapping causes at least two major issues: 1) users find and experiment with the old api which is now considered deprecated and unsupported, and 2) the existence of the old API is blocking the completion of the transition to bazel and causes other woes (like build times which exceed the time limit on the free instances of travis).
I still agree with two-years-ago-me, so if you think this is the right time, then I'm OK with it. I do like @jwnimmer-tri 's suggestion of adding a readme in the relevant deleted example folders, just so that things a new user might do (like a github repo search for "quadrotor" or just browsing to the examples folder) will still point them to the right SHA.