My first time participating in the All Japan Contest.
I have now been working on designing, building, assembling, writing code and testing Zeetah V (ZV) for over a year. While ZV can learn and solve my small five cell by six cell maze, it isn’t ready for the contest.
Having competed in Micromouse contests, off and on for many years, I’ve known that if I worked my butt off and was lucky, it would be at least a good two years before I would be competitive. However, I had hoped that by this time, ZV would be running well enough that it would not only learn and solve the maze but be able to do speed runs using diagonals in the path. Given the current status of ZV, I will be very happy if it learns the maze and makes it to the middle!
With time running out, my efforts had been very focused but I still hadn’t finished writing the longitudinal corrector for orthogonal or diagonals or the lateral corrector for diagonals. Just a day or two earlier, I had finally implemented a lateral corrector for orthogonal moves and it seemed to be working well but I hadn’t had a chance to gather data, analyze its performance and tune it.
Since I was flying out on Thursday morning, and was very tired, I contemplated stopping coding and packing for the trip. My wife suggested that instead of sleeping, I should pull an all-nighter. So, I stayed up and kept coding. By about 4AM, I had the longitudinal corrector working off of the side walls. I got up an hour before I had to fly out and packed and left for the airport.
The flight was uneventful. Since I was pretty tired, I didn’t write any code but I did think through what I wanted to do when I got to the maze since this would be the first time ZV would be in a full size maze. As the plane was on approach to Narita, I remember looking out the window and thinking about how crazy it was for me to fly 4700 miles to run a mouse in a maze. It didn’t help that ZV wasn’t ready. But I am glad to say that the feeling passed pretty quickly.
The flight got in forty five minutes before the bus to Tsukuba and the next one was two hours after that. I really wanted to make the earlier bus and fortunately, was able to get out of the plane and make it through Japanese immigration and customs in thirty five minutes. That gave me enough time to buy my bus ticket and walk over to the bus stop. I would have converted money at the airport but one of the guys at work had sold me ¥6000 he had left over from a previous trip and that with the ¥3000 I had from my last trip made me think that I didn’t need any more cash.
While waiting for the bus, I met up with Khiew Tzong Yong and his students who were taking the same bus. This made the hour and a half ride to Tsukuba go by really fast. Khiew Tzong is a wonderfully down to earth guy who is always willing to share info. During the course of our conversation, it became very clear that he works hard to keep things simple. For example, if you look at the acceleration profile for an integrated turn, assuming the traction limit for the tire is a circle, the differential velocity for the turn should follow a sinusoidal profile. Khiew Tzong, analyzed this and noticed that the difference between a trapezoidal profile and the sinusoidal profile is only at the transition points and tried the simpler trapezoidal profile. Guess what? He found the trapezoidal approach was giving him great results. How good you ask? At the 2010 All Japan Contest, he won the Half-size contest and came in third in the full size contest! The mice he did it with were less than a month old. If you get a chance, take a look at his Excel series of mice.
When I checked into the hotel, I found Peter had left me a note, so after I dropped off my stuff and freshened up, I met up with him and David Otten. Because of the time difference between our homes and Tsukuba, if you don’t want to wake up at 3AM, it is important to stay up as late as possible so that you are so tired that you sleep-in. So, we went for a walk. In the past, we’ve stopped at a crepe place for a snack but unfortunately, it is gone. So, this year, we ate ice cream.
Once I got back to the hotel room, I started to write more code. Eventually around 10PM, I went off to sleep. The staying up late trick didn’t work so well because I was up at 3:30AM.
I’ll post the next installment tomorrow.