Extremum-seeking control and applications. A numerical optimization-based approach (Q640680)

Extremum-Seeking Control (ESC) tracks a varying maximum or minimum in a performance function such as a cost. It attempts to determine the optimal performance of a control system as it operates, thereby reducing downtime and the need for system analysis. To place ESC in the context of the systems and control literature at large the authors note that ESC is highly related to both Traditional Optimal Control (TOC) and Model Predictive Control (MPC). In TOC problems, the performance function generally is an integral function of the state and control, and the extremum is a trajectory. Therefore, calculus of variations is involved in the design, and the analytical form of the performance function is needed to obtain necessary conditions of the optimal control. Related to optimal control, MPC is a widely applied advanced control method, where the control input at each step is obtained by solving an optimization problem. In this case, the optimization objective is to try to approximate an optimal trajectory in the context of optimal control using numerical optimization algorithms. The common thread between the three control approaches (TOC, MPC and ESC) is that they all try to solve the problem \(\min J(\cdot)\) subject to \(\dot x=f(x,u)\), where \(J\) is a cost function that may depend on states, control input, time, initial and final values. In order to solve this problem, both TOC and MPC methods generally assume that the function \(J\) is perfectly known. However, in ESC the performance function is generally a function of the state and is unknown or poorly known in the design -- and it is often a feature of the control problem itself. This lack of knowledge implies that one can only design an extremum seeking controller based on measurements of the performance function or its derivatives, if available. The monograph is divided into two parts. In the first, the authors review existing analog-optimization-based ESC including gradient-, perturbation- and sliding-mode-based control designs. They then propose a novel numerical-optimization-based ESC using optimization algorithms and state regulation. This control design is developed for simple linear time-invariant systems and then extended for a class of feedback linearizable nonlinear systems. The two main optimization algorithms -- line search and trust region methods -- are analyzed for robustness. Finite-time and asymptotic state regulators are put forward for linear and nonlinear systems respectively. Further design flexibility is achieved using the robustness results of the optimization algorithms and the asymptotic state regulator by which existing nonlinear adaptive control techniques can be introduced for robust design. The approach used is easier to implement than those that use perturbation-based ESC, and tends to display a high level of robustness. The second part of the book deals with applications of ESC: a comparative study of ESC schemes in antilock braking system design, impedance matching, source seeking, formation control, collision and obstacle avoidance for groups of autonomous agents, and mobile radar networks. The book will be of benefit to graduate students persuing their degrees in applied and industrial mathematics, as well as to researchers and engineers who may encounter optimal control methods in their work.