A 2 Month Research Trip to Osaka: FrontierLab@OsakaU

//Long Post Alert

My 2018 summer has been remarkably challenging and has kept me busy up until the very start of school. This includes my two day stay in Taiwan right after my last final during the Spring 2018 quarter ended, my two month stay in Osaka, Japan, studying rigorous algorithm engineering at the prestigious Osaka University, and a two week stay in Taiwan with my family that consisted of studying and software development that kept me up until roughly 4-5 AM daily. 

Finally, I had a final three week stay in my new apartment in Los Angeles, preparing for Operating Systems and Artificial Intelligence courses. Oh, I also spent time developing my GameMaker project with a team, working on an iOS application with my friend, developing two course curricula for SuperCoding, studying Cracking the Interview and Game Programming in C++, and wrapping up an online Computer Network Fundamentals course on Coursera. Furthermore, I spent around 120 hours in this time span developing my own site from a blank text document (which you can view here), which was an excellent learning experience. 

There has been nothing short of orderly chaos in my life over the last three months, and I cannot get enough of it! This is the kind of challenge I was expecting to go through at some point during my time as a college student. Regardless of the chaos, I must talk about my two month trip to Osaka, and how it has opened my eyes on the many wonderful elements of living in Japan that I have discovered first-hand.

I landed in Japan in the evening of June 17th, and walked straight into a massive earthquake that took place on the morning of the 18th. I was walking to the train station that would take me to Osaka University's orientation, but the train railway bridge above me made the loudest noise as the earthquake happened, and I honestly thought that it would fall and actually crush me. Luckily, it didn't (...). That was my real first 'welcome' to Japan. I lived in Kaohsiung, Taiwan, before moving to UCLA so earthquakes were not that big of a deal, but it was a nice opener.

After that, each day would consist of me researching something I had no clue about - the rotor router mechanism. Everything I found on the Internet pointed me to different directions (like a rotor router, haha), and I was lost for a good amount of time. My fluency in Japan led everyone (EVERYONE) in the lab to speak to me about everything in Japanese. This was fine, but hours of nonstop communication about topics ranging from men and women to computers and networks really trained my brain to process greater amounts of Japanese. 

I eventually got the handle on what the rotor router mechanism was about (which you can read about in my paper here), and started working on a big paper that looked at the benefits and disadvantages of using the mechanism. I was then suddenly bombarded with a midterm presentation report that I had to do in the lab sometime during Week 5 (the next week). Oh, and they wanted me to do it in Japanese - a new challenge that I was ready to take. You can check out the presentation slides here.

Okay, so after week 5, my work started to become more streamlined - I was able to write a lot about everything, and also completed the development of a C++ simulation program that would check the average stabilization period of a bot roaming a set-up graph. I will explain this in layman's term now (look at the graphic below): If we have a little robot exploring a bunch of computers (the circles), and the robot magically locks into a path that repeats forever. My simulation finds out how many steps (A to B is one step) it takes for the robot to lock into that repeating path, and what the average number of steps is over X number of simulations run on the same graph.

The results for one of the two graphs that I investigated (the graph above in this case) can be seen below. In a 2002 paper covering the basics of the rotor router mechanism, the stabilization period was proven to be O(2mD), which equates to 2 * the number of edges * the diameter of the graph. However, there has been no research on the actual stabilization period average for graphs, and this is where the importance of my research and paper comes in. The visual below clearly shows that despite the worst case for the (above) graph being 44, all of the million test cases led to the stabilization period being less than 13, and having an average of 1.25. This may have to do with the number of edges in the graph being high, but it still gives us a good look into the efficiency of the rotor router mechanism as a whole. 

I would love to talk more about the rotor router mechanism in a future post - particularly further enhancements that have been made to the original algorithm, and how using multiple robots at the same time would affect the stabilization period. If you have any questions about the mechanism, my paper, or anything else, definitely comment below or send me an email!