blown away extremely impressed by the speed of the top competitors during the recent MicroMouse competition in Tokyo. They weren't just fast, they were lightening fast, "faster than a speeding bullet" fast. I took a number of videos during the competition, and it was really hard to keep up with them. I had to use fixed focus because when I let my camera use it's auto focus function the video was blurred because the the focus couldn't keep up with the mouse.
There were, of course, slower competitors. Some of them just trundled along at their own pace. I'm not putting them down, not at all. The fact that they were able to navigate the maze is a major accomplishment by itself no matter how fast they were going.
When I first got interested in the MicroMouse competition I assumed that the major challenge would be exploring the maze and optimizing the best path. Naively I thought that the speed attempts would be interesting, but not necessarily the key to winning. After spending two days watching the actual competition and thinking about what I observed, I'm coming to the conclusion that the top competitors have very good exploration and analysis algorithms, and, more importantly, have spent an extreme amount of effort to shave hundreds of a second off their speed attempts. They are really what the Japanese refer to as 'takumi' - skilled craftsmen.
So, how did this evolve over time? That's going to take considerable research and analysis to figure out. The rules and the maze structure haven't undergone any major change during the 25 years that the competition has been taking place. The available technology has certainly improved very dramatically. And since the contest format has remained constant there was the opportunity for the accumulated knowledge to increase year after year. This may have been the most significant factor.
As time permits, I want to go back and look at the mouse design evolution over the years. For example, many of the mouse photo's I've found on the internet worldwide show designs with sensor arrays that extend over the walls and read the maze that way. Yet there were only a few of that type taking the field in Tokyo. All of the top competitors used design approaches that were down inside the maze.
I did manage to locate some of the performance statistics for the competition from its beginning in 1980 through last years (2003) event. At the initial event in 1980 no competitor was able to successfully complete the maze. In 1981 the winner navigated the course at an average speed of 189 mm/sec. Last year's winner managed an average speed of 1,609 mm/sec, getting very close to 10X the 1981 velocity.
I put together the chart below showing the way the winning speed (red) has increased over the years. I also included the path length in # of cells (green) and the number of turns (blue). Of course, the relative maze complexity is difficult to compare from year to year - at least I'm not sure how to tackle it yet. I'm also very curious about the drop off in speed from 1993 till 1997.
How much faster can they go? What new, innovative design approaches will be required? What is the asymptotic limit? Who will be the first to build a Mighty Mouse that can literally fly through the maze like Luke Skywalker flying in the canyons of the Death Star?
It's going to be fun to find out....
You might also enjoy: