The History of our sensors
The first sensor arrangement we tried was the middle sensor slightly out in front of the other two. This was a very good design for the linefollower problem, its only problem was that there were a small window of opportunity where the line would fall between the sensors. While, this did not cause problems for the completion of the line follower it did make its motion "choppy" as it bounced back an forth seeing and losing the line. This became more of a problem in the intersection finder problem as hitting the intersections evenly became important. If you were correcting your robot from being to far to the opposite side of a new branch, it was possible to miss this intersection. An example of this is consider going north and coming to a branch that looks like this |- if your robot is correcting itself from being to far to the left as it reaches the intersection, it will not realize there is a new branch going off to the right. So, smoothness is an important quality that can cut down on the number of intersections you skip over.

This led to trying to have the sensors all in a straight line, without spaces inbetween. The robot again moved smoothly on a straight line, but we were trying a new algorithm for intersection finder that removed the previous process of rotating 360 degrees everytime the robot thinks it found an intersection (for more on this algorithm see the comments in the actual code.) We were finding many false intersections, caused by sharp turns. This was due to the fact that too many sensors were on a line at a given time.

This led to the final configuration you see in the construction. The sensors are spaced out by one lego unit. This made the robot slightly less smooth, but this was countered by the fact that it was now harder to be considered an intersection. And the wider scope allowed us to see more of the area in front of the robot.