I'm so happy everyone is finding this article potentially useful. I apologize for "hiding" the rules table as a math formula instead of implementing it as a real table but the code runs twice as fast this way. It is important to remember the robot has no memory about previous moves; it only knows what adjacent squares look like now and makes a preprogrammed dna decision on what to do based on this current view. I could add some random walls in the grid and the robot would adapt and still perform very well.