My name is Isaac Krull and I'm a senior undergraduate student at Vassar College, graduating in the spring of 2009. I am a Computer Science major with a Mathematics correlate sequence and this summer I'm working and studying at Texas A&M University. The class I'm taking over the summer is Artificial Intelligence, a Computer Science class about intelligent searching using heuristics and learning algorithms. The research I'm doing with Professor Jennifer Walter, who is a tenured professor at Vassar College, is about motion planning for hexagonal self-reconfigurable metamorphic robots. Specifically, I've been tasked with designing, implementing and proving the correctness of a traversal algorithm which navigates pockets found in obstacle cells.
My experience here in Texas has been absolutely amazing thus far and I've already learned a substantial amount of information (including how to use HTML) so I can only imagine how great the rest of the summer will be. Now that my research is finally picking up steam and hopefully headed somewhere interesting, the days are going by faster and faster.
Week One: I spent the first week discovering the campus and familiarizing myself with the Parasol lab. I settled into my dorm nicely (didn't take more than a couple of minutes to completely unpack) and began to explore. The 'dorm' is more accurately described as a resort - it has its own pool, exercise facilities, theater, computer rooms and games. Everyone has been extremely friendly everywhere I go and a couple of the Computer Science REU students caught a playoff baseball game between UIC and Texas A&M. Everything is going great.
Week Two: The second week involved a whole lot more work. Deadlines began creeping up and the fact that I needed to learn HTML to even meet any of the deadlines meant that I was slogging through lesson after lesson online trying to figure it out. My research also began picking up steam: not only was I participating in a motion-planning crash course in robotics, but I had less papers to read and more thinking to do on my own - designing algorithms etc. Still, the living and working situations leave me nothing to complain about and I'm having a great time.
Week Three: My research this week hit a wall in the form of previously written code that just plain didn't work. The algorithm I was working on depended on function calls from the bridging algorithm that was developed before I got here and I discovered that the reason my program wasn't running properly was a problem with the smallest type of bridge. Therefore, I set about writing an algorithm to solve this issue, but got bogged down in the implementation because my solution depended on being able to alter the direction of rotation and substrate choices of individual modules. Once again, I had to abandon what I was working on to break it down into even smaller parts. I decided the first task I needed to accomplish was developing intermediate goals for the robots, so that they could progress from one goal formation into another. This would allow me for me to define the changes in substrates and directions by altering which goal modules were headed towards. I've completed a rough draft of the program for multiple goal configurations and have it working in a fragile state where only certain configurations can be handled. My work in the coming week will focus on making the program I wrote to do multiple goals much more robust and capable of handling goal configurations in any order on an obstacle surface.
Week Four: Having completed a more robust form of multiple goal configurations for the hexagonal robots but still unable to separate a single pocket of goal cells into multiple configurations my research focused on developing ways to identify conditions where a change of direction is warranted without using goal configurations to accomplish this. A comparison between perimeter values would be one way of accomplishing this task, so I reworked pre-existing perimeter code into the surface traversal script, marking the perimeter cells of each pocket independently. In the process though, I may have come across a more efficient solution by chance and will begin probing the possibility of classifying the direction change (to solve the infinite two-well) in terms of free sides which are pocket cells and free sides which do not have any attributes. I intend to attempt a rough implementation of the perimeter comparison in the week ahead also.
Week Five: This week, I developed an algorithm which identifies the three different bridge cases and applies the different required bridges to each. The simplest case can be handled using the existing code, so if it identifies a bridge of the BASIC type it simply allows the else statement to catch it and use the traversal code for walking over a surface. The other two cases, UP and DOWN bridges, can be identified by comparing the number of non-pocket, non-obstacle cells that surround the first and last cells in the bridge. If the first cell in the bridge has two free sides and the last cell has only one - then the bridge is a DOWN bridge. If the opposite is true, the first cell has one free side and the final cell has two free sides - then the bridge is an UP bridge. The implementation of the simulator code is going well, but the UP bridge timing is off when the pocket opening widens immediately after the mouth. A similar timing error has been encountered in the DOWN case, but it is working slightly better. In the next week I'll be addressing these weaknesses.
Week Six: Fourth of July! Oh yeah! This week marks a substantial success - not only do all three of the two module bridge cases work, but they work together (previously, I had been testing them separately and they were not happy together). Modules have passing over these bridges have been causing timing issues (possibly because the bridge modules are no longer being classified as FREE but I haven't worked that all out) but the majority of the work behind implementing my algorithm is done. I've begun the initial stages of my research paper (title, abstract and introduction) and my brother is coming down to celebrate Independence Day with me as well as take a tour of campus to see if he'd be interested in attending Texas A&M for his undergraduate degree.
Week Seven: My grandfather, William Krull, passed away this week. I did not advance my project very much and spent most of the week in Atlanta, GA with my family.
Week Eight: Tedious identification of special cases week! Pockets that are too shallow to use the full fledged form of my algorithm require special solutions. This also means that we have to go through and classify each case based on distinctive features. It appears that the most likely candidates to aid our identification task are the length of the perimeter, the number of obstacle cells adjacent to the first and last perimeter cells, places where the first perimeter cell is adjacent to the third perimeter cell and so forth. Demonstrating that the cases we've identified are the only cases appears to be another hassle on the horizon, but perhaps not overwhelmingly difficult.
Week Nine: The special cases for the algorithm have been identified, so now there are thirteen cases in all. With a good deal of help from Professor Walter, I was able to get some code up and running that handles each of these cases appropriately. In addition, this week saw the creation of my poster for the upcoming end-of-program presentations and poster sessions. Done in Powerpoint and featuring some nifty new diagrams depicting the work of the algorithm, I think it came out looking very sharp.
Week Ten: Things are winding down here. The physical print out of my poster has arrived, I'm finishing up the last struggles with my research paper and done with one poster presentation. Tomorrow we have two poster sessions to close out our program here and on Tuesday I have a final examination for my Artificial Intelligence class and then I'm out of here. Had myself a merry time but definitely found myself homesick a number of times throughout my experience here. With any luck, they'll accept my application to grad school here so I can make it back sometime.
Thanks for stopping by,
Parasol Home | Research | People | General info | Seminars | Resources
Parasol Laboratory, 425 Harvey R. Bright Bldg, 3112 TAMU, College Station, TX 77843-3112
email@example.com Phone 979.458.0722 Fax 979.458.0718
Department of Computer Science and Engineering | Dwight Look College of Engineering | Texas A&M University
Privacy statement: Computer Science and Engineering Engineering TAMU
Web Accessibility Policy and Law - Web Accessibility and Usability Standards - Contact Webmaster