Sharif Digital Repository

Using fractional order calculus to model complex phenomena with combined behavior (i.e., dissipative, capacitive and inertia behavior) is one of the most recent research fields in the world. These accurate models lead to better understanding of the phenomena. For example, consider viscoelastic materials. In early models, dissipative behavior is separated from capacitive behavior but by using fractional order models, these two behaviors are considered simultaneously thus the complexity of the model would be reduced. To use fractional order models in real-life engineering applications, it is important to investigate the dynamics of these systems. Unfortunately, there are a few proper... 

The constrained nonlinear systems with large operating regions have attracted great attention due to their correspondence with the most practical systems. There are several tools such as gain scheduling and Nonlinear Model Predictive Control (NMPC) to control them. Gain scheduling, with ability to provide stability guarantees between the estimated stability regions overlapping each other and to cover a large space of the allowable operating range of the system, is an attractive practical approach to control the systems with large operating regions. But this strategy do not account for constraints explicitly by online optimization. On the contrary, NMPC handles constraints on the manipulated... 

Disturbance is one of the inseparable components of the mechanical systems which cannot be avoided. In these systems a number of inner and outer sources exist which are the cause of disturbance. Abrupt changes in torque, uncertainty in parameters, mechanical impulses and external forces on robot’s parts all can be mentioned as examples which introduce disturbance that affects the output of mechanical and robotic systems. Therefore, disturbance rejection is considered indispensable in robotic control systems. There are number of problems which are associated with disturbance rejection. In several methods, mostly optimization based methods, system fails to completely reject the disturbance and... 

Many processes and real world systems indicate nonlinear behavior and operate either naturally in multiple regions or in a wide range of operation space. Modelling and control of these systems by nonlinear methods needs to have a deep insight. Furthermore, the nonlinear methods leads to complexity issues. To overcome these problems, multi-model methods are introduced which consist of two steps, decomposition and combination. In decomposition step, the complex system is divided into a set of simple models; in the combination step, the behavior of local models are combined. How to decompose to local models and how to combine them are two main questions in this area which not answered... 

Inthis thesis, we provide a reduced multiple linear approximationfornonlinear systems. This approximationis a weighted summation of reduced order models thatapproximates the states or output(s) of the nonlinear system. It can be used inreal-time control instead of high-ordernonlinear models in order toreduce thecomputational cost. The reduction of computation cost is claimed qualitatively, based on the linearity of the approximation and its less order compared to the original model. To form amultiple modelapproximation eachlinearmodelis obtained from linearization atanoperating point. Then,each model is reduced byalinearreduction method. We used truncation and balanced residualization... 

This project presents algorithms for intelligent control of the viral load in a HIV-1 infection model. The first part of thesis was dedicated to the study of the three states model adopted to describe the HIV-1 infection. A sensitivity analysis of the model parameters was described and computer simulations were provided to show the influence of the parameters explicitly. Then, an internal model control based on neural networks was implemented on the introduced infection model with including the RTIs and PIs drugs efficacies as control input. A stable adaptive neuro-control approach was presented for affine in the control nonlinear dynamical systems, whose nonlinearities were assumed to be... 

This thesis is concerned with stability and tracking of nonlinear noisy Lipschitz systems over the packet erasure channel with feedback acknowledgment when system and measurement are due to bounded noises. The desired stability and tracking criteria is bounded stability and tracking in probability. Two approaches are adopted to address the desired stability and tracking performance, one is based on the Chebyshev inequality and the other is based on the Binomial distribution. It is illustrated that when the erasure probability is large, the approach based on the Chebyshev inequality provides a better bound for stability and tracking in probability; while, when the erasure probability is... 

Design of optimal controllers for nonlinear systems and even for linear systems is one of the most important parts of the control science. In optimal controller design, our aim is to maximize or minimize some kind of cost function associated with that particular problem. Such cost functions are usually combining the overall control effort, system’s energy, or some other important specifications of the system. Explicit solution for linear quadratic controllers exists under some mild conditions such as stabilizability of the system. The solution is in the form of a full state-feedback law. But in practical problems, we may not have access to all of the system state variables. The problem can... 

Considering uncertainty is of great importance in the design of controllers for dynamical systems because the existence of uncertainty is inevitable in the modeling of almost all real-world systems. There are numerous methods and strategies for dealing with the control of uncertain systems in control systems theory including adaptive and robust approaches. The former one is more suitable for cases in which the information about uncertainty is very poor compared to robust techniques that are applicable in the cases more data is available about uncertainty. For instance, where there is no data about the direction in the actuator model, adaptive control methods are the only applicable choices....