Control of nonlinear dynamical systems. Methods and applications (Q938611)

This book has been devoted to the control of Lagrangian mechanical systems. It is assumed that the control algorithms should respect bounds on controls, and be robust to external perturbations and model uncertainties. The motion planning problem is studied, i.e. a control objective is to bring the system from a given initial state to a prescribed final state, possibly in a finite time. Two basic control synthesis methods are employed: the optimal control, and the stability theory. The presentation of the book is mathematical. Performance of presented algorithms is illustrated with numerous computer simulations of mechanical systems, including robotic systems. Specifically, a fully actuated Lagrangian control system \[ \frac{d}{d\,t}\frac{\partial T}{\partial \dot{q}}-\frac{\partial T}{\partial q}=U+Q \tag{lagr} \] is studied, defined by the kinetic energy \(T\) that takes the form of either a second-order polynomial in velocities \[ T(t,q,\dot{q})=\frac{1}{2}\dot{q}^TA(t,q)\dot{q}+a_1(t,q)\dot{q}+a_0(t,q) \] or of a quadratic form \[ T(q,\dot{q})=\frac{1}{2}\dot{q}^TA(q)\dot{q}, \] where \(U\) denotes control forces, and \(Q\) is a vector of bounded perturbations. The following control problem is addressed in the system (lagr): find a bounded feedback control \(U=U(t,q,\dot{q})\), \(| | U| | \leq U^0\), transferring the system from an initial position \(q(t_0)=q^0\) and velocity \(\dot{q}(t_0)=\dot{q}^0\) to a prescribed final \(q(t_*)=q^*\) and \(\dot{q}(t_0)=0\), at a certain time \(t_*<\infty.\) The book is divided into 10 chapters followed by a 123 item list of references and an index. A majority of the references are publications of the authors. The index is not very instructive. The chapters bear the following titles: Chapter 1 -- Optimal control, Chapter 2 -- Method of decomposition (the first approach), Chapter 3 -- Method of decomposition (the second approach), Chapter 4 -- Stability based control for Lagrangian mechanical systems, Chapter 5 -- Piecewise linear control for mechanical systems under uncertainty, Chapter 6 -- Continuous feedback control for mechanical systems under uncertainty, Chapter 7 -- Control in distributed parameter systems, Chapter 8 -- Control systems under complex constraints, Chapter 9 -- Optimal control problems under complex constraints, and Chapter 10 -- Time-optimal swing-up and damping feedback controls of a nonlinear pendulum. Below we make a quick overview of the book's content. Chapter 1 makes a statement of the optimal control problem, introduces the Pontryagin Maximum Principle, and presents analytic solutions to a couple of standard examples. Chapter 2 presents an optimal control synthesis based on a decomposition of the system (lagr) into independent linear, second order subsystems with perturbations estimated in terms of game theory. An extension to systems with velocity dependent damping is made. Chapter 3 develops further the decomposition and game theoretic approach. A trajectory tracking problem is addressed. Applications to robotics conclude the chapter. A short chapter 4 presents basics of the Lyapunov function method for autonomous and time-dependent systems. A modification to finite time stabilization is included. In chapter 5, a piecewise PD control algoritm of the form \[ U(q,\dot{q})=-\alpha\dot{q}-\beta q, \] with piecewise constant coefficients \(\alpha\) and \(\beta\), is designed, capable of steering the system (lagr) with unknown inertia matrix to its destination in a finite time. An estimate for the control time has been given. Convergence of the algorithm is proved using the Lyapunov function approach. Applications to several uncertain mechanical are shown. In chapter 6, a control algorithm of the similar form is devised, however, with coefficients depending continuously on time and phase variables. Its most general form is the following \[ U(t,q,\dot{q})= -\alpha(t,q,\dot{q})A_0(t,q)\dot{q}- \beta(t,q,\dot{q})q, \] where \(A_0(t,q)\) refers to the known part of the inertia matrix, and \(\alpha\) and \(\beta\) have been defined by means of a Lyapunov function. Chapter 7 makes an excursion into the area of infinite dimensional control systems described by partial differential equations of the form \(w_t=Aw+v\) or \(w_{tt}=Aw+v\), where \(w=w(t,x)\) is the state variable, \(A\) is a differential operator, and \(v\) denotes a control. The problem of steering these systems to \(w(\cdot,T)\equiv 0\) is studied, and solved using the separation of variables method and the eigenfunction expansion. An optimal control for resulting time-dependent expansion coefficients has been obtained. An application to the control of the elastic beam is considered. Chapter 8 deals with the control of time-dependent linear systems subject to instantaneous constraints of mixed type involving state variables, controls and some integrals of them. A standard, open loop control has been obtained, and sufficient conditions derived on the initial and final state of the system that guarantee the satisfaction of constraints. This control algorithm has been applied to a number of specific mechanical and electromechanical systems. Chapter 9 studies time optimal control of a second-order linear mechanical system, corresponding to the second Newton law, under various constraints imposed on its position, velocity or acceleration. The book closes with chapter 10, devoted to the synthesis of time optimal control for a pendulum whose terminal state is either the upper unstable or the lower stable position. To summarize, the book discovers a number of new and specific control theoretic ideas dedicated to mechanical systems that may arouse interest of control engineering community and provide a research inspiration to robotics community.