Matthew Powelson

@rkliman You don't have to create a new pull request. Just push to this branch and it will update.

It is close, we just need to restructure it a bit.

I have not reviewed it again, but if you think it looks ok that is fine. It looks like this includes the pagmo interface as well. Is that potentially breaking...

@arocchi You seem to be our resident optimization expert. Any thoughts?

@Levi-Armstrong I saw that. I guess I didn't fully understand what that meant, but I'm glad you're on top of it. To address some of the other questions we had,...

Yeah that looks nice. One negative that I see is that it is maintained by only one guy at ETH Zurich. I wonder if it will be around in a...

> [This project](https://github.com/locuslab/qpth) based on PyTorch may also be of interest. They implemented a batch QP solver (primal-dual interior-point method) which may be good enough for the use case of...

This is failing because it doesn't handle the collision buffer. I'll have to look more into what to do here. There is also interplay with the CollisionCoeffData

> @mpowelson what are the next steps on this PR? I'm not sure. We'll have to figure out how we want to handle the CollisionCoeffData and the Collision margin data....

If anyone goes looking, I've got the start of a pagmo interface [here](https://github.com/mpowelson/trajopt_ros/tree/feature/pagmo/trajopt_optimizers/trajopt_pagmo).