MIT Drake

drake.mit.edu

87 points·natcombs··8 comments

Having taken Tedrake's underactuated robotics class which uses Drake's python bindings, it is an amazing collection of tools and examples of algorithms. Really great for learning.

I would say is very focused on research and quick prototyping, however.

MIT Driverless experimented with using Drake in 2018-19 race car, but we found that just implementing everything in eigen was more flexible, removed a dependency, and more transparent for tinkering.

Ah the classic problem. Someday we'll get our shit together in Haskell land and just blow away these awful protype/production dicotomies.

Not if the smalltalkers get there first

Not if CPU's are fast enough to do it in Julia :-)

Ah, Julia shares my goals, but I really don't like it's methods.

To be we absolutely need static types and super high level stuff to both:

- keep the code closer to it's purpose

(e.g. do linear algebra and proof of why this program does the stats you want it to together)

- be super architecture agnostic

(With lambda generalized to arbitrary categorizes we can just elaborate to Vect and fuse functional pipelines with matrix multiplication).

(https://ncatlab.org/nlab/show/Vect)

"Model-based design and verification for robotics."

>Drake (“dragon” in Middle English) is a C++ toolbox started by the Robot Locomotion Group at the MIT Computer Science and Artificial Intelligence Lab (CSAIL).

I thought this was going to be https://www.factual.com/blog/introducing-drake-a-kind-of-mak...

Cmake project that depends on a specific distribution to be built _from source_? I know, it's 2020, we have Docker and things, but but but that's not how it works folks!

Reply on news.ycombinator.com