A strange occurrence

I recently wrote my way into an interesting bug as I was updating my python script for logging collisions.  The aim was to add in collisions for bins as well as netcode to get those bins from the web server.

Upon testing the new code I had produced, I found that some of the locations of the balls were getting set to 0 right before I tried to calculate whether they were colliding with a bin.  I ended up fixing this problem when I noticed that I was trying to collect the information on 5 bin types 30 times.  I only needed to talk to the server 5 times to determine where all the bins were, but it was getting this data 30 times, and continuing to append it onto a list.

Somewhere in this process, the over-sized list was causing a problem with the ball locations, and fixing the lists led to the disappearance of this bug.  I am still not altogether sure why this change fixed this problem, but for now, I have functional python behaving how I expect.

Almost Easy

The last couple of days I have spent trying to run a genetic algorithm to get feedback on how to tune the robots that will collaborate with the human.  These tests have not necessarily worked out.  The initial runs were using a remote server that has given us trouble in the past, and fared no better when I attempted it.  Today's attempts have taken significantly longer than expected, and have not even been completed, which is confounded by the fact that the data returned in the log files looks suspicious and maybe not meaningful.

I have also been working to improve the ability for the server to define the locations of target regions, and for the other programs to use this data to operate.  So far progress has been made, but has again stalled out due to what is most likely a very simple error in code logic in the javascript. As it stands now, though, the data coming in from the server is arriving correctly, but is being stored poorly, and all of the incoming data is being overwritten by the most recent data point, which obviously causes serious problems.

Making the same squares better

Today I've been working on completing the updates to allow the backend to handle the location of target regions.  It's interesting trying to determine the best system for all of the moving parts to communicate about this information consistently and make sure everything has all the information it needs.  In the end things will still look very similar but will allow us to make important changes much more easily for the experiment.  Unfortunately I am having a problem with the javascript that is contacting the server.  I am getting a 404 response code, but the same url when sent through my browser returns the expected data.

Paper with LaTeX

Last week I put together a paper and a presentation to go over the work I've done so far.  To write the paper, I used a really neat tool called LaTeX, which is like a small scripting language almost that allows for the formatting of documents.  With this tool, I was able to do a very minimal amount of work on format and simply link the file to a template created by acm which made everything appear nice and in its proper place.  I could also very easily change this template that is being used to have the paper show up as it would in any other format for other conferences etc.

Bringing it Together

I managed to meet with the psychologist, Dr. Pak, who will be running experiments on the platform we are developing.  We had a great meeting and were able to clearly define a lot of new objectives for the platform to help the experiment run as well as possible, with as much good data as possible.

We have got plans to improve the amount of interaction that participants will necessarily have, improve the clarity of the system to participants, increase participants' motivation to perform well, and increase the potential data we have to determine trust levels.

As we begin to implement some of these plans, we will be looking to get the experiment in a good place to be run early by Dr. Pak before we continue updating all of the possible pieces of the code.  Hopefully this will start giving us some very good results soon and meaningful data that we can use to help understand how people work with robots.