M.Ing.

Repetitive Control (RC) seeks to converge to zero tracking error of a feedback control system performing periodic command as time progresses, or to cancel the influence of a periodic disturbance as time progresses, by observing the error in the previous period. Iterative Learning Control (ILC) is similar, it aims to converge to zero tracking error of system repeatedly performing the same task, and also adjusting the command to the feedback controller each repetition based on the error in the previous repetition. Compared to the conventional feedback control design methods, RC and ILC improve the performance over repetitions, and both aiming at zero tracking error in the real world instead of in a mathematical model. Linear Model Predictive Control (LMPC) normally does not aim for zero tracking error following a desired trajectory, but aims to minimize a quadratic cost function to the prediction horizon, and then apply the first control action. Then repeat the process each time step. The usual quadratic cost is a trade-off function between tracking accuracy and control effort and hence is not asking for zero error. It is also not specialized to periodic command or periodic disturbance as RC is, but does require that one knows the future desired command up to the prediction horizon. The objective of this dissertation is to present various design schemes of improving the tracking performance in a control system based on ILC, RC and LMPC. The dissertation contains four major chapters. The first chapter studies the optimization of the design parameters, in particular as related to measurement noise, and the need of a cutoff filter when dealing with actuator limitations, robustness to model error. The results aim to guide the user in tuning the design parameters available when creating a repetitive control system. In the second chapter, we investigate how ILC laws can be converted for use in RC to improve performance. And robustification by adding control penalty in cost function is compared to use a frequency cutoff filter. The third chapter develops a method to create desired trajectories with a zero tracking interval without involving an unstable inverse solution. An easily implementable feedback version is created to optimize the same cost every time step from the current measured position. An ILC algorithm is also created to iteratively learn to give local zero error in the real world while using an imperfect model. This approach also gives a method to apply ILC to endpoint problem without specifying an arbitrary trajectory to follow to reach the endpoint. This creates a method for ILC to apply to such problems without asking for accurate tracking of a somewhat arbitrary trajectory to accomplish learning to reach the desired endpoint. The last chapter outlines a set of uses for a stable inverse in control applications, including Linear Model Predictive Control (LMPC), and LMPC applied to Repetitive Control (RC-LMPC), and a generalized form of a one-step ahead control. An important characteristic is that this approach has the property of converging to zero tracking error in a small number of time steps, which is finite time convergence instead of asymptotic convergence as time tends to infinity.

prepared by John E. Ward, Mark E. Connelly, Avram K. Tetewsky. / Final report / Bibliography: p. 188-193. / "July 31, 1978." -- "September, 1978."--Cover. "Submitted to: Transportation Systems Center, Department of Transportation, Kendall Square, Cambridge, MA 02142." / DOT-TSC-1274 M.I.T. Project. 84492

Multiple Hypothesis Tracking (MHT) is generally the preferred data association technique for tracking targets in clutter and with missed detections due to its increased accuracy over conventional single-scan techniques such as Nearest Neighbor (NN) and Probabilistic Data Association (PDA). However, this improved accuracy comes at the price of greater complexity. Sequential K-best MHT is a simple implementation of MHT that attempts to achieve the accuracy of multiple hypothesis tracking with some of the simplicity of single-frame methods. Our first major objective is to determine under what general conditions Sequential K-best data association is preferable to Probabilistic Data Association. Both methods are implemented for a single-target, single-sensor scenario in two spatial dimensions. Using the track loss ratio as our primary performance metric, we compare the two methods under varying false alarm densities and missed-detection probabilities. Upon implementing a single-target Sequential K-best MHT tracker, a fundamental problem was observed in which the tracks coalesce. The second major thrust of this research is to compare different approaches to resolve this issue. Several methods to detect track coalescence, mostly based on the Mahalanobis and Kullback-Leibler distances, are presented and compared.

No description available.

No description available.

Traditional Automatic Target Recognition (ATR) Systems often fail when faced with complex recognition tasks involving noise, clutter, and complexity. This work is concerned with implementing a real time, vision based ATR system using an Artificial Neural Network (ANN) to overcome some of the shortcomings of traditional ATR systems. The key issues of this work are vision, pattern recognition and artificial neural networks. The ANN presented in this thesis is inspired by Prof. Stephen Grossberg's work in Adaptive Resonance Theory (ART) and neurophysiological data on the primate brain. An ANN known as Selective Attention Adaptive Resonance Theory (SAART) (Lozo, 1995, 1997) forms the basis of this work. SAART, which is based on Grossberg's ART, models the higher levels of visual processing in the primate brain to provide an ATR system capable of learning and recognising targets in cluttered and complex backgrounds. This thesis contributes an extension to the SAART model to allow a degree of tolerance to imperfections including distortion, changes in size, orientation, or position. In addition to this extension, it is also demonstrated how modulated neural layers can be used for image filtering. A possible extension of the architecture for multi-sensory environments is proposed as a foundation for future research. / Thesis (MEng)--University of South Australia, 2000

Traditional Automatic Target Recognition (ATR) Systems often fail when faced with complex recognition tasks involving noise, clutter, and complexity. This work is concerned with implementing a real time, vision based ATR system using an Artificial Neural Network (ANN) to overcome some of the shortcomings of traditional ATR systems. The key issues of this work are vision, pattern recognition and artificial neural networks. The ANN presented in this thesis is inspired by Prof. Stephen Grossberg's work in Adaptive Resonance Theory (ART) and neurophysiological data on the primate brain. An ANN known as Selective Attention Adaptive Resonance Theory (SAART) (Lozo, 1995, 1997) forms the basis of this work. SAART, which is based on Grossberg's ART, models the higher levels of visual processing in the primate brain to provide an ATR system capable of learning and recognising targets in cluttered and complex backgrounds. This thesis contributes an extension to the SAART model to allow a degree of tolerance to imperfections including distortion, changes in size, orientation, or position. In addition to this extension, it is also demonstrated how modulated neural layers can be used for image filtering. A possible extension of the architecture for multi-sensory environments is proposed as a foundation for future research. / Thesis (MEng)--University of South Australia, 2000

Thesis (Ph. D.)--University of Washington, 2007. / Vita. Includes bibliographical references (p. 111-114).

Thesis (Ph. D.)--Ohio State University, 2007. / Title from first page of PDF file. Includes bibliographical references (p. 108-112).