Existence of solutions and existence of optimal solutions. The quasilinear case (Q1066070)

Using Kakutani type fixed point theorems and monotone operator theory in new ways, the authors obtain existence theorems for abstract operator equations which include quasi-linear evolution equations and optimal control problems involving Cauchy-Dirichlet and Cauchy-Neumann equations. Specifically, the authors obtain existence of optimal solutions without requiring seminormality for Mayer type and Lagrange type problems monitored by (1) \(Ex(t)+(\Lambda x)(t)=g(t,(Mx)(t),u(t))\), u(t)\(\in \omega (t)\), \(t\in J=[0,T]\), \(x\in X=L_ p(J,V)\) where V is a separable reflexive Banach space with dual \(V^*\). Assuming that \(<\Lambda v,v>\geq \alpha \| v\|\) and \(\| g(t,z,u(t))\| \leq \ell (t)\) for measurable selections u(t) of \(\omega\) (t), the authors show sequential weak continuity of \(N_ u(x)=\Lambda x-g(\cdot,Mx,u)\) and show that \(<N_ ux,x>\geq 0\) for \(\| x\| \leq (\| \ell \| /\alpha)^{1/(p-1)}.\) An existence theorem of Hou for solutions of \(Ex+N_ ux=0\) yields the existence of solutions for (1) in \(K=\{x| \| x\| \leq R\}\) where \(R\geq (\| \ell \| /\alpha)^{1/(p-1)}\). Using a variant of the McShane-Warfield implicit function theorem, the class \(\Phi\) of all solutions of (1) for all possible measurable selections u of \(\omega\), is shown to be sequentially weakly compact. These results are then applied to the case where \(E=(d/dt)\) with domain \(D=\{x\in X\), \(dx/dt\in X^*\), \(x(0)=0\}.\) Using the maximal monotonicity of E on \(X=L_ 2(J,V)\) and Hou's recent result that maximal monotone operators on Fréchet spaces have Cesari's property (Q), the authors obtain existence of weak solutions of the abstract Cauchy problem dx/dt \(\Lambda\) x\(=f\) with \(x(0)=0\), \(f\in X^*=X=L_ 2(J,V)\). Several special cases are then considered. These existence theorems are then applied to obtain optimal solutions for Mayer and Lagrange type control problems. Using a fixed point theorem of Hou for multifunctions \(\Gamma\) : \(K\to K\) with closed graph, controllability is proved for problems monitored by \(Ex(t)=g(t,Mx(t),u(t))\), u(t)\(\in \omega (t,Mx(t))\) for \(t\in [0,T]\). Several illustrative examples are given.