The main contribution of this thesis is the formulation of a new type of resonant mode controller for which a closed form expression is derived. This controller requires no design iterations and the gain margin can be specified ahead of time (phase margin is always infinite). This controller differs from standard resonant control in that it acts directly on a disturbance rather than damping resonant poles of the system plant. This allows for a notch to be created in the disturbance anywhere in the spectrum, not just at resonant modes of the system. The experimental setup involves headrest and headset active noise control systems using speakers and microphones. Experimental results using an adaptive version of the controller show large reductions (\u3e 20 dB) are achievable, even for frequency varying disturbances. A simple LQG weight synthesis algorithm and a procedure for improving the SOCIT system identification software are also given.
展开▼
