Paying for Amazon Web Services at CSU

Patsy said there is some kind of continuing payment form she can fill out for our AWS purchase and she will do that today.

Update: We need Jin in the dean's office to help us with this but she's not back till Wednesday. Website will have to wait...

Grand Opening

We had our lab Grand Opening on July 19th and it was a great success. The President of CSU, the Dean of Fenn Engineering College, our department chair, and the VP of Research at Parker Hannifin were all there and most spoke. Ton gave a nice speech and iterated the importance of academic freedom in the relationship with Parker Hannifin, which was a nice touch. After the speeches and ribbon cutting Ton, Sandy, and Obinna gave a nice demo to the crowd of our lab's capabilities. I really enjoyed the demo of treadmill feedback control based on moving a marker around in the motion capture space. It was like a wizard of oz moment with Ton commanding the powers of nature. We then had a nice reception down in the lobby with free booze and food (the veg options were ok but the stacked all the sausage together with the cheeses which was sorta gross). Photos of the event are here:

https://plus.google.com/photos/110966557175293116547/albums/5906047607418526609

Benefits

I think I'll go with Metrohealth select because I don't go to the doctor much. But the locations don't seem to be to convenient to where I live. It is a 9.9 mile bike ride to the main medical center, but the Buckeye center is probably only 2.5 miles and they have an allergy department which I may want to use.

Still wondering whether I should chose the Ohio state retirement deal or choose something on my own that isn't tied to the pension system in Ohio. Waiting on a call back from Angela Moss to discuss.

Moving Expenses

Yumi and I spent about $2,950 bucks to move from Davis to Cleveland. The Amtrak tickets (sleeper car from Davis to Chicago, coach Chicago to Cleveland) were $1150, shipping about 2300 lbs of boxes via Amtrak was about $1400, we rented a truck for about $100 on the Davis side to take our boxes to the train station, and I gave my brother $200 for driving up from Danville with his jeep to help us move the boxes on the Cleveland side, and the rest was some packing materials. I asked for moving assistance when I got hired and Ton agreed to pay $2k, so I guess we came out all right. The Amtrak trip was super fun (barring the Sacramento station fiasco, which I'll write about later).

These were submitted and are waiting the return of Jenny on Wednesday to get the account numbers for the reimbursement.

Lab Printer

I downloaded the CUPS drivers for our Xerox 6280N printer from here:

http://ftp.xerox.com/support/phaser-6280/downloads/enus.html?operatingSystem=linux&fileLanguage=en

Extract the tar file, then the rpm, and then dig into the folders to get Xerox_Phaser_6280N.ppd.gz and ungzip it.

We have a two tray setup with 256mb of memory (I think). Not sure if there is a duplexer.

Structural Model Identification

I worked on the structural model id last week. The due date for the conference paper is the 31st of July. I updated the code to deal with multiple runs per plant and it now saves a csv file with the meta results of the identifications. One major problem was that I wiped my computer and reinstalled everything. This was great except I no longer had a working copy of Matlab 2010a. I managed to get a copy of the 2013a to work with but the System ID toolbox had a major revision and were were now getting different results. The noise estimations were no longer converging for the simple plants and the results just didn't match what I got when running the code on 2010a. WTF!? This is why working with closed source crap like Matlab is a nightmare. Why would the results change from one version to the next? Do they not use unit tests!! Probably not, since the unit test framework was only added in the 2013a version. Anyhow, Ron took the results and finished up the paper. There is no way to verify this either if you don't own multiple versions of Matlab.

I've realized that lumping the plant into the closed loop A matrix was a bad idea for the noise estimation. In our case we know the plant because it is generated by our simulink code, we need K=0 for anything to do with the plant. We also know the input (error in the signal) and the output (stick control) for each run. I should be ID'ing just on that data. Ron asked me why I wasn't doing that a long time ago and it didn't dawn on me. The same goes for my dissertation work on the bicycle, I should redo that analysis too and maybe get some decent noise estimations.

Here is the updated code, data, and paper:

• Source code: https://github.com/moorepants/structuralid
• Data: http://figshare.com/articles/SISO_Human_Operator_Tracking_Tasks/96176
• Paper

In my opinion, this still needs some major work. I haven't really been happy with the results but Ron seems to think it is ok. Maybe I can rethink it all and turn it into something better. We really wanted to id the transition in control gains over time given a switch to a different plant but haven't had much success with that yet.