The complementary-slackness class of hybrid systems (Q2565306): Difference between revisions

The term ``hybrid systems'' applies to systems which have both continuous and discrete components. For example a mechanical system which is mathematically represented by a continuum and works in a continuous environment, may be coupled with a digital computer. Formulations of early papers on such hybrid systems contained studies of dynamical systems that contain both continuous and discrete variables. Recent interest in such systems arose in connection with studies of controls for digital computers interacting with continua. This occurs in studies of computers controlling chemical processes, in robotics, in studies of mechanical systems subjected to a bang-bang type control, etc. Because this field is so large and relatively new, there exist serious language problems, and even more basic problems of what kind of mathematical modeling is most appropriate for such systems. The most common approach uses differential-algebraic mathematical models (DAEs), which in the authors' presentation take the form: \(\dot x_i(t)=f_i(x_i(t),u(t))\). Here \(x_i(t)\) represents a continuous state of a dynamical system, attached to a node \(i\) in a graph of discrete events. A continuous output of such a system is given by the equation: \(y(t)=h_i(x_i(t))\). The times of transitions are determined by the state at the node \(i\) and by a discrete input. The constraints on the hybrid system are generally algebraic. The simplest such systems contain only equations with constant coefficients and no forcing functions. The authors refer to a system of equations \(f(z(t),\dot z(t))=0\) as the fully implicit form. If there is no input, such system is called autonomous. This use of the term autonomous must not be confused with the usual use in the theory of differential equations, where it denotes an absence of explicit dependence of the state on time. To offer an example of jump or impulsive solutions, the authors offer a simple example of two carts connected to each other and to a rigid wall by elastic springs. Motion of one of the carts is restricted by a rigid stop. The algebraic inequalities arising in this example resemble those in optimization problems, which are known as ``complementary slackness'' conditions. The authors proceed to study linear complementary slackness systems of the type: \[ \begin{aligned} \dot x &= Ax(t)+Bu(t),\;x\in\mathbb{R}^n,\;u\in\mathbb{R}^k,\\ y &= Cy(t)+Du(t),\;y\in\mathbb{R}^k;\;y(t)\geq 0,\;u(t)\geq 0,\;y(t)^Tu(t)=0.\end{aligned} \] For bimodal systems, the authors offer the following lemma: The system is autonomous if and only if the transfer function is not zero, or equivalently if not all Markov parameters vanish. Moreover, if the leading Markov parameter is positive then the complementary slackness equations are well-posed. A detailed discussion of Hamiltonian dynamics in a specific mode follows in the next section of this paper. In the discussion of jump points, a fundamental projection rule for the Hamiltonian is proved. The authors refer to it as the energy inequality. Simple mechanical examples are offered, including a lengthy analysis of constrained motion of a pendulum suspended by a weightless rope. They certainly help to assimilate the fairly sophisticated ideas advanced by the authors. While hybrid differential-algebraic systems have been subject of studies by several authors, such as H. S. Witsenhausen and Werner Rhinebold in the late 1960-s or 1970-s, the literature listed by the authors illustrates the sudden growth of interest in such systems in the late 1980-s and almost an explosive growth of research at the present time. The article under review is an important contribution to this new field connecting mathematical control theory to theoretical computer science, particularly automata theory.