Our central design principal was stability. We strived from the beginning to avoid complicated encoder and feedback schemes for our positioning, instead favoring systems that had physical limitions on their range of motion. This created a system that relied little on extensive calibration to deliver the chips with reasonable reliability. This helped with the cost constraints on the project, allowing us success using the cheaper, class-provided motors and servos without sacrificing accuracy.
Our design also involved construction out of lightweight, and aesthetically pleasing laser-cut acrylic, and standard aluminum angle and bar stock, arranged to allow maximum mechanism visibility. We kept our machine footprint small and also kept the total system very lightweight. It can easily be moved by one person. The relative low-cost of these materials allowed us to iterate through design several times when necessary to optimize part fit and geometry. While this was a new medium for our mechanical engineers, they adapted to it quickly, and ended up rather enjoying the simplicity of the medium, which can move quickly from design through fabrication.
The rest of this page will detail the subsystems roughly in the order the chips pass through them, from feeder through deployment into the board, followed by a look at the electronics (including software) behind it all. Diagrams accompany the description of each subsystem. These are simplified for the purpose of exposition of the basic concepts.
Chips are dropped into the hopper at the top of the machine. When the new game command is issued, the motor running the agitator (Fig. 1) is turned on, which orients the chips properly to enter the queue (Fig. 2). The agitator is turned off once fourteen of the twenty-one chips are queued in the ramp system, and then will be periodically activated again when there is space in the queue for more. If a chip gets jammed in the agitator, a reflectance sensor detects the lack of rotation and will reverse to free the jam with a very high success rate. At the end of the ramp system there is a servo-controlled actuator (Fig. 3) that can release a single chip into the chip dropper below. The servo used for this is a standard analog servo (model Hitec HS-325) with two arms on it. The top arm is extended to prevent the remaining chips in the queue from prematurely falling into the chip dropper, while the bottom arm is withdrawn to allow the bottom chip to pass.
|Fig. 1 - Agitator Wheel|
|Fig. 2 - Chip Queue|
|Fig. 3 - Single Chip Release Mechanism|
Immediately upon receiving the opponents previous move, a single chip enters the chip dropper (Fig. 4), which is deployed into the neutral zone as described below. The chip dropper begins with a chip release mechanism (Fig. 4 - top right) very similar to the one on the chip feeder, but with only the one arm since there is no need to worry about multiple chips in the dropper (even if somehow more than one was released, there is no room in the mechanism for multiple chips). Once the AI determines the best move, the single chip is released by a micro servo (model Hitec HS-81) into the final ramp (Fig. 4 - bottom two-thirds) where the column selectors deliver it to the proper column. The column is selected by three acrylic assemblies attached to three servos (again, model Hitec HS-325). Each servo controls access to two of the columns, and the last column is always open as a fail-safe. The stability of the system here is ensured by loose tolerances and a physical limitation of motion in one direction. The column-selection servo is activated before the chip is released to make sure that it is in the proper position before the chip reaches it. The chip rolls down the ramps until it reaches the open column, where it falls down into the selected column.
|Fig. 4 - Chip Dropper|
|Fig. 5 - Chip Dropper - CAD Render|
Neutral Zone Entry
Our neutral zone entry is unique among the teams in the class. Instead of depending on linear actuation, such as a belt, rack and pinion, or lead screw assembly, to reach the column, we enter from behind the board with a four-bar linkage system. Our entire dropper assembly advances into the neutral zone and sits above the game board. This gives us the advantage of a very simple and precise method of positioning the dropper over the board. The alignment is guaranteed because the assembly is at its physical limit in both the extended and retracted states. As long as our baseplate is aligned properly on the studs, the probability of a miss is quite low.
|Fig. 6 - Neutral Zone Entry - Side view|
|Fig. 7 - Neutral Zone Entry - Arm retracted|
|Fig. 8 - Neutral Zone Entry - Arm Extended|
Lastly, we will briefly explain our unique communication system. Though we use the prescribed 9600 baud 8N2 encoded serial communication protocol, our exact implementation is unique. It is unique in that instead of bidirectional connections between the different boards and controlling PC, we have a one-way loop through the system. This was to both save output pins, and take advantage of the fact that there is only one buffered serial port on the Arduino boards we used. The communication arrives from the controlling PC to the Pro, which chooses a move and issues commands to the Uno. Once the Uno has carried out the move, it reports the move back to the controlling PC.
TeamOur team is comprised of four students, two of whom come from a mechanical engineering background, and two from an electrical engineering background. Below is a little information on each member, including which portion of this project they primarily focused on.
Jon Daneman is a senior mechanical engineering student. His main area of work for this project was designing and fabricating the chip feeder subsystem, and lots of other CAD work, such as generating animations for progress reports.
Anton Galkin is a graduate mechanical engineering student. His focus was design and fabrication of the chip dropper subsystem, as well as machining and assembly of many components, camerawork (both photo and video), CAD modeling and rendering.
Greg MacLean is a senior electrical & computer & biomedical engineering student. His main area of work for this project has been designing and implementing the game AI that the system uses to play Connect Four. He also did video editing for our final video. In his free time, Greg studies combat strategies to teach Jormungandr how to more effectively defeat Thor.
Warren Pryde is a senior electrical & computer engineering student. For this project, Warren's main work was in implementing the control software to run the actuators and process the sensor inputs. He also handled wiring, sensor/actuator calibration, and system integration, in addition to making the website.
VideoBelow is the video we produced for our final presentation detailing the subsystems and showing the machine in action.