Dynamic simulation of controls in certain parabolic systems (Q1180881): Difference between revisions

Let \(U\) and \(E\) be two real Hilbert spaces, \(\varphi: E\to(- \infty,+\infty]\) be a convex, proper, lower semicontinuous function, \(A=- \partial\varphi\), \(B: U\to E\) be a linear bounded operator, \(f(.)\in L_ 2([t_ 0,T];E)\) and \(P\subset U\) be a convex, bounded, closed set. The problem of reconstructing the unknown control \(u(.)\in L_ 2([t_ 0,T];E)\), \(u(t)\in P\) a.e. in \([t_ 0,T]\), in the nonlinear parabolic inclusion \[ \dot y(t)\in Ay(t)+Bu(t)+f(t);\;t_ 0\leq t\leq T,\;y(t)\in E,\;y(t_ 0)=y_ 0 \] is considered. The reconstruction of \(u(t)\) must be done on the basis of the measurements \(\{\psi_ i\}^ m_{i=0}\) which are known to satisfy \(\|\psi_ i-y(\tau_ i)\|_ E\leq h\), \((h>0\), given), where \(\tau_ i=t_ 0+i(T-t_ 0)/m\), \(i=0,1,\dots,m\). An algorithm for constructing approximations of \(u(t)\) based on the feedback principle is proposed. It is indicated that under suitable assumption on the modulus of continuity of \(y(t)\), these approximations tend in the \(L_ 2([t_ 0,T];U)\)-norm to the unique unknown control with minimal \(L_ 2([t_ 0,T];U)\)-norm, as \(h\) tends to zero. The same approach is applied to boundary control and distributed control problems for linear parabolic equations. Illustrative examples are given.