Theory and implementation of robust performance digital servo controllers.

摘要

The first controller structure is a model reference controller of the state feedback type. The controller employs past observations of the system's response to obtain an estimate of a function representing the effect of system perturbations. The gathered information is subsequently used to cancel the unknown dynamics and unexpected disturbances. An important feature of the controller is that precise knowledge of the functional form of the uncertain dynamics is not required. This leads to a control structure with a model independent behavior. A tradeoff requirement for good performance is that the dynamics stemming from the perturbations should not have a rapidly varying behavior which degrades the estimation process. Within this requirement, the need for knowledge on upper bounds of parameter variations and disturbances is eliminated. The feedback control gains can be substantially reduced leading to improved tracking performance.; In this dissertation sampled data models of continuous-time systems are expressed in terms of an incremental operator called the 'delta' operator. The advantage of this approach, over a traditional shift operator approach, is three-fold. First, sampled data models whose limiting properties match those of the underlying continuous-time system can be constructed. Second, the relationship between discrete-time and continuous-time domain analyses can be made more transparent if the 'delta' operator is used. Third, and most importantly, restructuring sampled data models in terms of the 'delta' operator provides a possibility of handling sampling zeros.; A method proposed in this dissertation is to restructure sampled-data systems in terms of the 'delta' operator. By using the 'delta' operator, when the sampling interval is regarded as a parameter, the zero dynamics of a sampled system is shown to be singularly perturbed i.e. exhibits two-time scales. The implication is that, with fast sampling, the sampling zeros tend to infinity while the rest of the zeros tend as a set to the zeros of the underlying continuous-time system. Consequently, unstable sampling zeros may be ignored if the sampling interval is selected to be sufficiently small. Attached experimental tests verify, that while ignoring the zeros introduces an error in the discrete-time model, the resulting control laws give satisfactory performances.; The second proposed controller structure is similar to the first but requires only output feedback. It is based on the idea of the disturbance observer which generates estimates of an equivalent disturbance acting on the system. In turn, the gathered estimate is used as a cancellation signal to make the plant output follow a desired reference signal. The dissertation proposes a direct discrete-time design approach to the disturbance observer. It is shown that the disturbance observer formulated in the discrete-time domain yields better performance than its counterpart developed in the continuous-time setting.; The proposed controllers are shown to improve the robustness to parameter variations and disturbance suppression of motion control systems. Computer simulation and experimental tests on a direct drive dc-motor positioning system, a two-degree of freedom SCARA robot and a high-speed xy table designed for micropositioning are provided to corroborate the theoretical developments. (Abstract shortened by UMI.)