SUBJECT: Ph.D. Dissertation Defense
   
BY: Daichi Fujioka
   
TIME: Wednesday, June 1, 2016, 2:00 p.m.
   
PLACE: MARC Building, 114
   
TITLE: Input-Shaped Model Reference Control for Flexible Systems
   
COMMITTEE: Dr. William Singhose, Chair (ME)
Dr. Charles Ume (ME)
Dr. Jun Ueda (ME)
Dr. Joshua Vaughan (ME)
Dr. Jonathan Rogers (ME)
 

SUMMARY

Heavy-lifting machines, such as cranes, degrade their operation efficiency due to the inherent flexible dynamics of the systems. The problem is further complicated by complex nonlinear dynamics, time-varying parameters, and lack of the full state information. This thesis investigates on developing a simple and robust control method that improves the operation of flexible systems even in the absence of an accurate system model and sensing. This goal is achieved via the combination of input shaping and model reference control. The benefits of the proposed controller design include the increased shaper robustness against plant uncertainties and parameter estimation errors, while achieving good vibration suppression and control effort reduction. The input-shaped model reference control scheme is presented. The optimized controller design is developed to minimize the time-delay in the controller while assuring large robustness to errors. The proposed controller design is tested on single- and double-pendulum payload cranes. The system dynamics are explained and the state space representation of the reference model and the plant are derived. The Lyapunov control law with asymptotic stability requirement is derived to formulate the control signal. Analysis reveals that input-shaped model reference control contributes to reducing the control effort magnitude for large ranges of system parameter values and the parameter variances. The robustness of the proposed controller in state tracking and oscillation suppression performances are analyzed and verified via numerical simulations and experiments. The performance of the proposed controller design is validated further by conducting a human operator testing on obstacle course navigation.