ENHANCED SOFT COMPUTING TECHNIQUES FOR ROBUST CONTROLLER DESIGN

Abstract

Last few decades have seen the rapid advances and diffusion of technologies in science, engineering, robotics, biomedical, economics and other fields. Diverse technologies are incorporated in industrial processes to increase the efficiency, which in turn alleviated the complexity of the systems. Highly specialized skills and expert knowledge are required to design, develop, operate, and control such complex systems. Along with this, the uncertainties caused by inaccurate modeling, external disturbances, and variations of working conditions influence the system performance. Eventually, design and development of robust control system with intelligent computational tools is essentially required to get desired performance in efficient manner. In recent years, the concept of soft computing based intelligent control has emerged as an efficient tool to enhance the existing non-linear, optimal, adaptive, and stochastic control methods. The intelligent control can be achieved through the involvement of various artificial intelligence (AI) and soft computing approaches utilized for closed-loop feedback control to improve system performance, reliability, and efficiency. The major soft computing techniques are fuzzy logic system, metaheuristic algorithms (MAs), chaos theory, neurocomputing, and probabilistic reasoning. Various soft computing techniques and their fusions are commonly used to enhance the intelligent control tools by incorporating human expert knowledge in computing processes. The field of soft computing is always growing by contributions from the large community of researchers and provides an exceptional opportunity to advance its methodology and applications. Consequently, there is a great scope and motivation to ameliorate, design, hybridize, and apply these techniques. This fact motivates us to present some significant improvements and novel contributions to major components of soft computing like fuzzy logic system and MAs. The main aim of the overall work presented in this thesis is to enhance the soft computing techniques to improve the performance of the control system design. The complete work in this thesis is distinguished by four research objectives, given as (a) Design efficient optimization algorithm for enhancing the performance of complex systems, (b) Performance analysis of the proposed algorithm for optimization of different controller design problems, (c) Design of interval type-2 fuzzy precompensated PID (IT2FP-PID) controller applied to 2-link robotic manipulator with variable payload, and (d) Constrained multi-objective optimization (MOO) approach for robust controller design and performance analysis. The tuning of controllers is considered as a high-dimensional, complex, multimodal numerical optimization problem, as many locally optimal solutions can be obtained for different combinations of the parameter values. Thus, it is always a challenging task for designers to get the optimal tuning parameters. MAs are extensively considered for solving such control system design problems to get the best performance and robust response. Some common deficiencies faced by the majority of population-based MAs are lack of exploration ability, slow and premature convergence behaviour, and stagnation to local optima. Considering the above-stated problems, important features of two well established MAs, grey wolf optimizer (GWO) and artificial bee colony algorithm (ABC), are hybridized to develop an improved GWO-ABC algorithm as the first objective of the thesis. In GWO-ABC algorithm, information sharing property of employed bees in ABC is adopted with conventional GWO algorithm to comprehend the benefits of both the algorithms. A new population initialization strategy is introduced to get widespread range solutions. These strategies are incorporated to overcome the shortcomings of the conventional GWO algorithm by improving exploration capability, convergence rate, and reduce the chances of entrapment at local optima. The performance of theGWO-ABC algorithm is validated through extensive experimental analysis of 27 benchmark functions and comparisons with 5 other standard intelligent algorithms. After successfully substantiating that GWO-ABC algorithm is an efficient algorithm for solving complex test functions, it is applied to control system design problem for linear and non-linear test bench process plants. Here, the GWO-ABC algorithm is applied to minimize the objective function so that optimal time-domain specifications could be achieved. All the design requirements like low overshoot, better rise time, faster settling time, minimum steady-state error, and performance index are evaluated and compared to other state-of-the-art algorithms. Further, the conventional GWO algorithm is improved by incorporating the communication signalling strategy used in cooperative foraging of wolves. The leadership hierarchy approach and communicating behaviours are merged to present improved cooperative foraging based GWO (CFGWO) algorithm. New acceleration coefficient is proposed to balance the exploration and exploitation behaviour throughout the iterations. The proposed algorithm is examined on a real-world optimization problem of controller designing for trajectory tracking problems of a 2-link robotic manipulator with payload at tip. The comparative graphs of trajectory tracking performance, the path traced by the end-effector, and X and Y coordinate versus time variations against their desired reference curves are plotted. Also, the plots of position errors and controller output for both the links are also presented. As mentioned earlier, many of the real-world industrial processes are influenced by large amounts of uncertainties due to dynamic unstructured environments. The type-2 fuzzy logic controllers with type-2 fuzzy sets are highly recognized to deliver a satisfactory performance in the face of uncertainty and imprecision than their type-1 counterparts. In order to establish its applicability, an efficient IT2FP-PID controller is presented for trajectory tracking of a 2-link robotic manipulator with variable payload. The controller is comprised of unique features of interval type-2 fuzzy precompensated controller cascaded with a conventional PID controller. The fuzzy logic controller (FLC) based precompensator is incorporated to regulate the control signal to compensate the undershoots and overshoots in the system output when the system has unknown non-linearities. Tuning of control parameters and the antecedent MF structures emerged as a complex, high-dimensional, and constrained optimization problem. Hence, a systematic strategy for optimizing the controller parameters along with scaling factors and the antecedent MF parameters for minimization of performance metric integral time absolute error (ITAE) is viii presented. The structures of MFs are also optimized to get maximum benefits of footprint of uncertainty (FOU) in type-2 FLC. Prominently, GWO-ABC algorithm is utilized for solving this high-dimensional constrained optimization problem. In order to witness effectiveness, the performance is compared with type-1 fuzzy precompensated PID (T1FP-PID), fuzzy PID (FPID), and conventional PID controllers. The efficacy of the proposed controller is also validated through exhaustive robustness analysis in presence of distinct non-linear dynamics such as (i) payload variations, (ii) model uncertainties, (iii) disturbance in signals, and (iv) random noise at feedback path. After experimental outcome, it is inferred that IT2FP-PID controller outperforms others and can be adopted as a viable alternative for controlling non-linear complex systems with higher uncertainties. As a whole, the work in this objective manifests that (a) additional tuning parameters provide extra degree of freedom (DOF) to get better performance in optimal controller design, (b) in case of IT2-FLC, the systematic strategy to optimize the shapes of MFs derive maximum benefit of FOU to handle uncertainty (c) the proposed IT2FP-PID controller revealed as viable alternative to control complex non-linear systems with high uncertainties, (d) GWO-ABC algorithm can efficiently solve the low- and high-dimensional constrained optimization problems. The last part of the thesis is dedicated to investigate the applicability of MOO approach for tuning of controller parameters for multi-variable, constrained, and complex control systems. In the control system design, minimization of performance error indices for set-point and disturbance are considered as conflicting objective functions with various constraints. A simple design and parameter tuning strategy for 2-DOF fractional order PID (2-DOF-FOPID) controller is presented using MOO approach. A fast and elitist non-dominated sorting genetic algorithm (NSGA-II) with constraint handling methodology is utilized and the sensitivity function is used as a constraint. The major robustness investigations are carried for minimization of integrated absolute error (IAE) for both set-point tracking and external disturbance rejection. The comparative performance evaluation is assessed against equivalent counterparts and found that the proposed 2-DOF-FOPID controller performs with superior results.