The goal of this project is to establish and control formation movement of an EvBot II colony without explicitly stating which position each EvBot II is to maintain. Each EvBot II will compete for its position in the formation based on an internal heuristic that is a function of: (1) the location of the EvBot II, (2) the desired formation, (3) the proximity of the EvBot II to the formation configuration, and (4) an internal self-diagnostic control algorithm.
The positions within each formation will be ranked from most to least desirable, but the heuristic will be such that a robot that is already near a formation position will compete more fiercely for that position than for any other position. The self-diagnostic algorithm is designed to mimic real world applications where problems could arise that would make the mobile robot less competitive in the formation, such as running low on fuel or taking damage. This heuristic will allow for dynamic changes in the formation while the colony is completing a task. A mobile robot that begins to run low on fuel will begin to lose its competition for its spot and will start to be demoted through the formation. The hope is that if the robot does cease to function then it is already toward the back of the formation where it will hinder the fewest number of other mobile robots in the formation. Conversely, if this algorithm were being used in a military application a damaged robot may be demoted such that it migrates into the center of the formation where it is more protected from enemy fire.
The formations will be maintained through a vision based system where a webcam is pointed upward at a spherical mirror. This gives each robot a distorted three-hundred and sixty degree field of view. Each robot is a different color and knows which color it is so that once the positions are decided a vision recognition algorithm searches the webcam image for the appropriate colors. The EvBot can then determine the angle and the distance to each of the robots in its field of view. These robot locations will be feed directly into the control algorithm to maintain the correct position within the formation. This vision based system allows for greater flexibility since the mobile robots can operate without knowing their global positions in the test world.
Faculty Team: Edward Grant
Phone: 919-515-7016
Emails: daburke@ncsu.edu; egrant@ncsu.edu
