Notebook Entry

AIO for March 10, 2015

Last week's accomplishments

Quiet Standing Identification Paper

  • [0.6 hr] Worked on packaging opty for conda.
  • [4.9 hr] Tried to animate the SVG with matplotlib and worked on the PyDy 3D animation.
  • [3.7 hr] Specified the acceleration inputs more precisely based on some other papers.
  • [2.6 hr] Worked on getting the scaling to work with the new acceleration input.

Gait Control Direct ID Paper

  • [4.1 hr] Got the repository running and updated all source code to GATK 1.2.0.
  • [5.5 hr] Started on the draft and made pretty good head way.
  • [0.5] Worked a bit on speeding up the inverse dynamics.

Yeadon Software Paper

  • [1.4 hr] Finished up all tasks for the second revision and now waiting on the typesetters to publish.

Last week's issues

Perturbed Data Paper

I didn't get to any of the issues this past week. I'll be finishing that up this coming week now that the deadline is here. There is only one issue that may take some time. One reviewer requested that I update the meta data notes for all of the trials with extra information. I'm not sure I need to do that. Any thoughts on it would be appreciated:

https://github.com/csu-hmc/perturbed-data-paper/issues/134

Quiet Standing Identification Paper

I did not finish the first draft because I've been hung up on dealing with problems that arose when introducing a "better" platform input specification. I got hung up, so I worked on the gait direct id paper instead. I still would like some feedback on the issues I've posted about this.

The problem this week stemmed from changing the platform acceleration input to a slightly different sum of sines. I use a 12 sine wave logarithmically spaced in frequency with random phase shifts from 0 to 2 pi and a linearly decreasing amplitude wrt to frequency. This is not much different than my original input but IPOPT doesn't converge as easily nor as quickly with an initial of all zeros (it does converge when the initial guess is the measure states and zeros for the gains). I'm thinking to move away from an all zero initial guess as that was originally in place just because it worked. In practice you will always have some estimate of the state to use in an initial guess.

This week's objectives

Perturbed Data Paper

  • [15 hr] Close all remaining issues.
  • [0.5 hr] Have Ton and Sandy review the changes and the response to the reviewers.

Quiet Standing Control Identification Paper

  • [4 hr] Work further on the new input and scaling and adjust code.
  • [2 hr] Create plot of relative error vs dc step size.

Gait Control Direct ID Paper

  • [10 hr] Get all of the basic plots in place in the paper.
  • [5 hr] Flesh out the results writing more for a first draft.

Yeadon Software Paper

  • [0.5 hr] Paste in the reviewer responses into the commenting section on the journal website.

Zhoadan Collaboration

  • [2 hr] Review Google Research grant reqs and intro materials.
comments powered by Disqus