Team members:
- Kyle Huston
- Dan Mackin
- Farhad Majdeteimouri
- Ahmad Muhammad
Our project will use self-organizing maps (SOM) to model a strand of optical fiber. Essentially, what this entails is shining laser light through a multi-mode fiber-optic strand and recording the resulting speckle pattern using a CCD sensor. The Matlab Image Acquisition Toolbox will then be used to pull this pattern into Matlab. The user will frequently tap the fiber at different points and the SOM algorithm will eventually learn the topology of the fiber. After that, analysis of the data will determine the exact location of where the user is tapping on the fiber. Once we have developed this system, we will pursue further expansions in dimensionality and speed.
Current systems use a definite number of sensors to detect movement on a structure. Our system implements a string of "sensors" (neurons) that is potentially less costly that current setups, both in construction and expansion. Because of the number of taps (samples) needed to learn the topology, we intend to automate the sampling process utilizing a microcontroller to both position and control the tapping mechanism. We could use the idea to model three dimensional non-rigid body motion for use in the aerospace and robotics industry.