Notebook Entry

Today's task list:

  • [x] Review Tarun's and Sachin's PRs
  • [x] Work on the website content and structure
  • [] Work on BMD papers
  • [~] Do CITI course
  • [x] Work on parsing the walking data
  • [x] Submit PyCon proposals
  • [] Working on wrapping Ton's walking models
  • [] Book hotel for BMD
  • [] Post update about BMD copyright
  • [] Finish reading the van der Kooij paper
  • [] See if our controller can drive an OpenSim model or Ton's 2D model
  • [] Wrap the HBM C code
  • [] Duplicate website backups on a S3 bucket
  • [] Work on the website theme
  • [] Make generic settings on the lab website
  • [] Review the TODO items on the Yeadon paper
  • [] Do FERPA course, due Sept 20
  • [] Write up database proposal
  • [] Try out CSympy with some mechanics problems
  • [] Email Mounir about teaching

System Identification

I've been using sys id for a long time now and learnt it on my own for my dissertation work, but I've hit enough walls in just being a user that I really need to start understanding what the hell is going on. I like Ljung's book "System Identification: Theory for the User", but it think the title would better be "System Identification: Theory for the Theorist" because he doesn't show you how to actually implement things computationally, which is what a "user" really needs. So I'm going to work through his book from page one and come up with the computations for the main methods in a series of IPython notebooks. I hope that my knowledge will then be sufficient to design a general system id package in Python and actually be readable (unlike the system id toolbox in Matlab).

The repo is at:

And the computations for Chapter 1 have been started here:

Walking System Identification

Over the weekend I implemented a method to zero out the identification of particular gains in the gain matrix (closing issue

The results are here for the example data:

I'm not sure why the sum of the residuals is being returned as zero from numpu.linalg.lstsq. It seems like the A matrix becomes rank deficient once I delete columns which correspond to different feedback loops. It also seems like the gains that I don't zero out have similar profiles to the previous full gain matrix calculations.

I also found some bugs and spent a long time working on it today. Will report more tomorrow.

PYCON 2014

I submitted a PyDy tutorial and a python for human walking data analysis for PyCon 2014 in Montreal. We'll see what happens...