Last week's accomplishments

Objectives from last week:

Inverted Pendulum Parameter Estimation Source

• [4 hr, Done took 6 hr] Implement general time specific constraints, like x(0) = 0.
  • See this commit: https://github.com/csu-hmc/inverted-pendulum-sys-id/commit/7face7257f025fbc8b9f95c6005ceb9f87b1e048
• [1 hr, Done took 2 hr] Add Ton's pendulum example to the code.
  • The software can now solve general direct collocation problems with minimal problem setup and a nice API. This shows how you'd solve the same Problem Ton presented during the HMC seminar but with lots less coding: https://github.com/csu-hmc/inverted-pendulum-sys-id/blob/master/pendulum_swing_up.py
• [1 hr, Done took 3 hr] Add the pendulum parameter id example.
  • This example shows a simple parameter identification. Note that the backward Euler integration introduces damping in the solution but get's near the parameter result. Once I implement the trapezoidal rule, this will converge better. See this file for the implementation: https://github.com/csu-hmc/inverted-pendulum-sys-id/blob/master/vyasarayani2011.py
• [5 hr, Done took 0.5 hr] Decide what to do about process and measurement noise and implement it.
  • See this discussion: https://github.com/csu-hmc/inverted-pendulum-sys-id/issues/7, still up in the air.

Perturbed Walking Data Paper

• [2 hr, Done took 1 hr] Read the Betts paper more thoroughly.
  • This paper implements the parameter id methods that I have done, except in a more robust, efficient (except maybe the numerical differences part), and general way. So I don't think what we have is fundamentally new, but it may be new for the biomechanics world, especially if I can identify the walking control.
• [3 hr] Decide on what will actually be the processed data and write script to produce it from raw to prepare for a Zenodo upload.
  • I think I'm only going to host the raw data along with the GaitAnalysisToolkit which can parse it and not host any processed data. I wrote a script to package the raw data for zenodo or figshare: https://github.com/moorepants/walking-sys-id/blob/master/src/prep_data_for_zenodo.py
• [6 hr, Done took 0 hr] Add in the belt acceleration compensation to GTK.o
  • This is nixed for now. Ton pointed out that since we've done the identification on the unperturbed data and get the same results, then this is likely inconsequential.

Bicycle Steer Torque Paper

• [3 hr, Done took 1 hr] Change R plots to a prettier ggplot2 version.

Bicycle Data Paper

• [1 hr, Done took 1 hr] Make a figshare data set for the additional trial csv file.
  • http://dx.doi.org/10.6084/m9.figshare.1187092

IEEE Open Source Code Night

• [2 hr, Done took 1.5 hr] Prep materials.

Other stuff I did:

• [3.5 hr] Gave the second IEEE Open Source Code night tutorial.
• [7 hr] Working on adding the Ackermann et. al 2010 walking open loop identification problem as an example to the direct collocation software.
• [1 hr] Updated DTK to work with a range of dep versions and fixed a bug.
• [1 hr] Updated BicycleParameters to work with latest DTK and pinned versions.
• [1.8 hr] Pushed two PRs to sympy to fix issues associated with differentiating the walking model.
• [1 hr] Discuss steer torque paper with Mont.

Last week's issues

Inverted Pendulum Parameter Estimation Paper

Didn't get to this besides reading the Bett's paper more thoroughly. It will be my main reference, I believe.

• [3 hr] Fluff up the intro with citations in the outline and rewrite it to discuss the issues with shooting and the direct approach.

Still trying to clarify the goals of the inverted pendulum paper and what to actually include.

Bicycle Steer Torque Paper

I'm working to update the software packages that are used to create the matplotlib plots so that I can run everything as a complete pipeline with the latest dependencies. So the matplotlib improvements are holding off till then.

• [2 hr] Improve the matplotlib plots to make them prettier.
• [1 hr] Figure out the best citation for Kageyama's design.

Bicycle Data Paper

Didn't get to this.

• [1 hr] Make a figshare item for the bicycle parameter raw data.

This week's objectives

Walking System ID

• [1 hr] Read Manoj's recent paper.

Inverted Pendulum Paper

• [0.25 hr] Email Samin to inquire about co-authorship.
• [1 hr] Discuss paper direction with Ton.
• [1 hr] Rework the outline to reflect Ton and my discussion.
• [1 hr] Implement the first example in the Betts et.al 2003 paper.
• [2 hr] Implement the Kuo standing model for the paper.
• [0.5 hr] Get the numerical values of the human model to match Kuo.
• [0.5 hr] Obtain examples of process noise levels that will produce nominal wiggly human motion.
• [4 hr] Implement process noise, measurement noise (both on states and torques), and perturbation accelerations.
• [5 hr] Finish implementing the Ackermann 2010 example.
• [3 hr] Get two relevant sympy prs into sympy.

Walking Data Paper

• [0.5 hr] Decide what will and will not be published data.
• [1 hr] Describe the GaitAnalysisToolkit DFlow data and it's output.
• [1 hr] Decide on what plots to show that display the nature of the data and add to the outline.
• [1 hr] Write a closing paragraph.

Bicycle Steer Torque Paper

• [4 hr] Update BicycleParameters to work with the latest Yeadon and uncertainties 1 and 2.
• [2 hr] Use seaborn to improve the matplotlib plots.
• [2 hr] Go through Mont's review, fix minor stuff, and prioritize larger changes.

Bicycle Data Paper

• [3 hr] Update BicycleDataProcessor to work with latest DTK, BicycleParameters, and PyTables.
• [0.5 hr] Create a raw data H5 file for upload to Zenodo.
• [1 hr] Create a raw and processed data H5 file (no filtering) for upload to Zenodo.

Other

• [2 hr] Send ENG 4 teaching materials to Yuri.