Non-linear robust observers for systems with non-collocated sensors and actuators.

摘要

Challenges in controlling highly nonlinear systems are not limited to the development of sophisticated control algorithms that are tolerant to significant modeling imprecision and external disturbances. Additional challenges stem from the implementation of the control algorithm such as the availability of the state variables needed for the computation of the control signals, and the adverse effects induced by non-collocated sensors and actuators.;The present work investigates the adverse effects of non-collocated sensors and actuators on the phase characteristics of flexible structures and the ensuing implications on the performance of structural controllers. Two closed-loop systems are considered and their phase angle contours have been generated as functions of the normalized sensor location and the excitation frequency. These contours were instrumental in the development of remedial actions for rendering structural controllers immune to the detrimental effects of non-collocated sensors and actuators.;Moreover, the current work has focused on providing experimental validation for the robust performances of a self-tuning observer and a sliding mode observer. The observers are designed based on the variable structure systems theory and the self-tuning fuzzy logic scheme. Their robustness and self-tuning characteristics allow one to use an imprecise model of the system and eliminate the need for the extensive tuning associated with a fixed rule-based expert fuzzy inference system. The first phase of the experimental work was conducted in a controlled environment on a flexible spherical robotic manipulator whose natural frequencies are configuration-dependent. Both controllers have yielded accurate estimates of the required state variables in spite of significant modeling imprecision.;The observers were also tested under a completely uncontrolled environment, which involves a 16-ft boat operating in open-water under different sea states. Such an experimental work necessitates the development of a supervisory control algorithm to perform PTP tasks, prescribed throttle arm and steering tasks, surge speed and heading tracking tasks, or recovery maneuvers. This system has been implemented herein to perform prescribed throttle arm and steering control tasks based on estimated rather than measured state variables. These experiments served to validate the observers in a completely uncontrolled environment and proved their viability as reliable techniques for providing accurate estimates for the required state variables.