Optimum \(H^ \infty\) designs under sampled state measurements (Q1175509)

The author considers the following \(H^ \infty\)-optimal control problem. The time interval is \([0,t_ g]\) and the system is described by the vector differential equation \[ \dot x=A(t)x+B(t)u(t)+D(t)w(t), x(0)=0, \] where \(U(t)=\mu(t,x(t_ 0),x(t_ 1),\dots,x(t_ k))\), \(t_ k\leq t<t_{k+1}\); \(\mu\) is piecewise continuous in its arguments, \(t_ 0,\dots,t_{K-1}\) is such that \(0=t_ 0<t_ 1<\dots<t_{K-1}<t_ g\). \(M_ s\) denotes the set of all such controllers (\(S\) for sampled). Finally \(W_{[0,t_ g]}\) denotes the \(m\)-dimensional disturbance, taken to be square-integrable on \([0,t_ g]\); \(H_ w\) denotes the Hilbert space of such functions, with \(\| w\|^ 2=\int_ 0^{t_ g}w^ T(t)w(t)dt\). The quadratic cost function is: \[ L(u,w)=x(t_ g)^ TQ_ gx(t_ g)+\int_ 0^{t_ g}[x(t)^ TQ(t)x(t)+u(t)^ Tu(t)]dt. \] For each fixed \(\mu\in M_ s\), \(L\) induces a natural mapping of \(H_ w\) on to the positive reals, denoted by \(G_ \mu\). The minimax (\(H^ \infty\)- optimal) controller design problem is to find a controller \(\mu\) which minimizes \[ g(\mu)=\sup_{w\in H_ w}\{(G_ \mu(w))^{1/2}/\| w\|\}. \] A complete solution is then obtained to this problem; in a second part the limiting behaviors of the finite-horizon \(H^ \infty\)- optimal controller \(\mu^ s\), and the attenuation bound \(\gamma^ s=\inf_{\mu\in M_ s}g(\mu)\) as \(t_ g\to+\infty\) are studied. Finally an example is provided to illustrate the results in both finite and infinite horizyon. The basic approach is game theoretic, by associating a particular differential game, and is characterized in term of the solution of a particular generalized Riccati differential equation, with the optimal performance determined by the conjugate point conditions associated with a family of generalized Riccati differential equations.