A robot capable of identifying puzzle pieces and manipulating them to complete the puzzle.
Individual Project, 7 months
For my senior thesis at Princeton, I chose to design and develop a robot capable of solving jigsaw puzzles. While it is true that this has applications in artifact reconstruction and crime scene forensics, I chose this project because it was fun and recognizable. Everybody has seen jigsaw puzzles before but very few have seen a robot complete one. It is surprisingly difficult to get a machine to perform a task that many children can do easily. I chose a 3x4 puzzle because it can be further generalized to any size, and because I wanted the robot to complete the border first, as a human would.
First, the pieces were separated and the robot traversed the area, taking pictures of each piece with a camera. These images were sent to a computer to solve the puzzle digitally and determine where each piece needed to be. The pieces were picked up and released using a suction cup connected to a small vacuum pump. This worked for retrieving and depositing individual pieces, but I noticed that when humans put puzzle pieces together, they use one hand to hold a piece and the other to keep the adjacent pieces in place. Adding springs with rubber feet allowed me to replicate this feature while solving the same issue. This part of the robot is shown in the pictures below.
The piece picker moves throughout the available area by using tracks on each side of the grid. A belt moves when the motor rotates a pulley, causing the wheels to slide along the track. There is a similar track to move the piece picker across the middle of the grid. This is similar to the motion of a CNC machine or 3-D printer. More details of the sliding carriage can be seen in the illustrations below.
The CAD model and hardware assembly are shown below, along with a wiring diagram. Overall, the robot performed well, although there is room for improvement with the precision of the camera as well as the motor movement. Fitting the puzzle pieces together requires very tight tolerances, as the puzzle will not look nearly as good if there are any gaps or overlapping pieces. In the case of overlapping pieces, it would become much more difficult to reassess a piece's new position and lift it up while remaining true to the camera's view of the piece.
Comments