Matt shows that the bike is fully tested and working while Storch saws down the bolts holding the motor in place.

Results and Reflections

Our demonstrations were successful and the finished project met the original design goals. The bicycle shifted up and down appropriately and held the current gear when the rider was pedaling in the specified cadence range. See the system in action here. (Thanks to Peng for the video!)

We used the following parts in the creation of our project. All of the parts mentioned were available in the lab or created in the machine shop, so our project budget was essentially zero.

• Pittman GM9413-5 motor
• Potentiometer (knob-type, single turn)
• Aluminum spindle with set screws
• Derailleur cable (on bike)
• Panasonic DN6848 Hall effect sensor
• National Semiconductor LM1875 power operational amplifier
• Two-pole switch
• Various resistors and wire
• Various bolts, screws, and metal plates to mount the motor
• Heat shrink and nylon "zip" ties
• Epoxy
• Neodymium magnet
• Project board with 12 and 5 volt available power supply
• +/- 15 volt power supply
• PC104 stack

We also came up with several improvements that could be used by future groups who wish to work on this problem. The most obvious extension is to add this functionality to the front gear cog as well.

One thing that we had a lot of trouble with was determining what gear the bicycle was actually in. While we got the motor feedback to work for us, it was not without its difficulties and correctly determining the current gear required hours of testing. If this could be directly measured by a sensor on the gear cog itself, it would be more reliable and simplify the programming task.

Another thing that could be added is a means of "filtering out" backwards pedaling. Right now, pedaling backwards fools the system and the behavior is the same as if the rider were pedaling in the proper direction. However, if the transmission attempts to shift while the rider is doing this, the chain can fall off the cog. This could be accomplished by using a pair of Hall effect sensors in close proximity to each other. Programming logic could be used to determine which way the user was pedaling by looking at which sensor was activated "first" in a rotation.

A third development to add would be support for coasting, that is, the time when the rider is not actively pedaling. Right now, the system can handle it if the rider ceases pedaling within their cadence range, but will still attempt to shift if the rider coasts after entering either shifting range. It was very difficult to differentiate "no pedaling" from "very slow, but still pedaling."

One final goal for future development would allow the rider to set their own desired cadence and shift ranges. Riders often determine this in conjunction with the gear ratios on their bike. Ideally, they could set a different range for each gear. A fingertip control pad and display would also be useful elements in such a design.