Summer 2005
Working with Dr. Dirk Grunwald and Dr. Greg Grudic I spent Summer 2005 developing a robot to detect wireless internet signals.
We used the nearest neighbor algorithm to predict specific locations where the robot would be. Our research showed that wireless signals bounce too much for precise location.
Through this research I bumped into many obstacles: the robot's direction was skewed on carpet, developing the interface, and analyzing the loads of data I received.
I programmed the robot to talk with people from remote text commands. In addition the robot could sing and dance which was fun to do while collecting data to unsuspecting prospective students.
The above data shows what my robot analyzed. Each one of these was one wireless stations strength, with over 200 actually being analyzed. As you can see the variablity of strength made the data fairly inconsistant. If I were to continue on this experiment, I would attempt to use more sophisticated machine learning algorithms (perhaps Randomized Forrests) to deal with the chaos of the data.
Working with Dr. Dirk Grunwald and Dr. Greg Grudic I spent Summer 2005 developing a robot to detect wireless internet signals.
- recieved $4,500 NSF grant for undergraduate research
- developed a planning and control system for the ER1 robot
- collected data to be used with a Nearest Neighbor Algorithm.
- helped show how individuals could be tracked through wireless internet signal strength
We used the nearest neighbor algorithm to predict specific locations where the robot would be. Our research showed that wireless signals bounce too much for precise location.
Through this research I bumped into many obstacles: the robot's direction was skewed on carpet, developing the interface, and analyzing the loads of data I received.
I programmed the robot to talk with people from remote text commands. In addition the robot could sing and dance which was fun to do while collecting data to unsuspecting prospective students.
The above data shows what my robot analyzed. Each one of these was one wireless stations strength, with over 200 actually being analyzed. As you can see the variablity of strength made the data fairly inconsistant. If I were to continue on this experiment, I would attempt to use more sophisticated machine learning algorithms (perhaps Randomized Forrests) to deal with the chaos of the data.