I achieved the majority of my goals either entirely or to some limited degree. My overarching goal of creating a robot that could map mazes and solve them by traversing the shortest path was only partially realized. I was able to implement a more basic algorithm, the left wall follower. However, this algorithm only allowed the robot to solve mazes without loops and lacked any mapping capacity. As a result, the robot could complete the maze when placed in very specific starting locations, but it could not complete the maze from the true starting point. Unfortunately I was unable to successfully implement the flood fill code to the 3pi. I was successful in keeping the cost below $150, with an estimate of around $132. More than half of this cost stemmed from the standard 3pi robot. Had I designed my own robot, I certainly would have been able to decrease the total cost even further. I also accomplished all of my goals regarding maze design and construction. I was satisfied with the complexity of my final maze design, and although I knew it would be difficult to create a perfectly accurate sized and neat maze, my work on that was as much as I could have hoped for. The installation of functional encoders onto the 3pi was an additional success. This seemed to be a choke point in terms of project completion for other similar projects that I found online, so I am glad that I was able to get them working. Although I was not able to extensively test the encoders with maze run throughs, they provided ideal output readings when hooked up to an oscilloscope.

There are various improvements and additions that could be made for this project. The addition of the encoders introduced some issues with motor values for turning because the wheels were pushed further out from the robot body. Also, there was a lack of digital I/O lines on the 3pi itself, so I could only attach one encoder, in this case, the left wheel encoder. This affected my ability to take accurate encoder readings. The Arduino software also posed a challenge as it was difficult to debug my code. The flood fill algorithm could also be improved upon since it returns the shortest path distance wise, but not necessarily time wise. My maze design, however, would not have encountered this type of problem even had the code been functioning. Some other improvements that could be made include implementing flood fill first in Java and then translating into Arduino so I could debug beforehand in the jGrasp software, and create my own robot so that I could have more I/O lines available for customization.