I mentioned Itani-san's MM3 website in my previous post. He has included a program he wrote that simulates different maze search strategies. First you either load the maze you want to test from a file, or build and save it interactively. To test the program I built the maze used for the APEC 2002 contest. Once you have the maze setup, you click on the magnifying glass icon and are prompted to set the parameters for the search. In the webpage text he explains the primary options and what the trade offs are for each of them. Then you press a button to run the program.
A second window opens and you can see the maze as it looks to the mouse as it does its search. In the figure the maze appears in the left window, and the mouse with its path in the second. The mouse is the yellow figure and the green dots represent cells it has already visited. It also projects a shortest path to the goal indicated by the white path. When I was watching the competition this past weekend I was really impressed by the fact that many of the mice seemed to guess the correct path very quickly, almost as if they already knew which way to go. Now, after studying Itani-san's program, I'm beginning to understand how they achieved their magic.
In this screenshot the mouse has already completed enough of its search to know that it has found the shortest path. It is retracing it with the path in white with yellow symbols showing the mouse in each of the cells in the path.
When the program finishes it displays the optimum path in blue in the left window, the cells visited in green in the right window, and reports the total number of steps taken, number of steps in the optimum path, and the number of turns.