SUBJECT: Ph.D. Dissertation Defense
BY: Matthew Marshall
TIME: Thursday, June 27, 2013, 12:00 p.m.
PLACE: Love Building, 210
TITLE: Weighted Multi-Camera Uncalibrated Visual Servoing Using a Decentralized Adaptive Kalman Filter
COMMITTEE: Dr. Harvey Lipkin, Chair (ME)
Dr. Nader Sadegh (ME)
Dr. Jun Ueda (ME)
Dr. Ayanna Howard (ECE)
Dr. Wayne Daley (GTRI)
Dr. Ai-Ping Hu (GTRI)


This thesis introduces a control method for image-based robot guidance that exploits the advantages of multiple cameras. It provides system survivability in the event of image occlusion or camera failure and produces a control action based on statistically meaningful data weighting without any prior knowledge of robot or camera parameters. The research presents a novel control law that uses a Kalman filter and can track a moving target. Adaptive filtering improves filter performance and it is decentralized for robustness and distributed processing. System stability is shown and the control method is supported experimentally and by simulation.