Optimal control and admissible relaxation for subdifferential evolution inclusions (Q2563950)

The paper is devoted to the solution of two optimal control problems for dynamic systems described by the differential inclusions of subdifferential type, namely: The original problem (OP) \[ -\dot x(t) \in \partial \varphi (t,x(t))+\int_K f(t,x(t),u(t),z) \mu(dz),\quad t \in T=[0,b], \] \[ x(0) = x_0, \quad \mu \in M_+^1 (K),\quad u(\cdot)\in S_U, \] \[ m = \inf_{u\in S_U} \sup_{\mu \in M_+^1 (K)} \int_0^b L(t,x(\mu,u)(t),u(t))dt, \] and the relaxed one (RP) \[ -\dot x(t) \in \partial \varphi (t, x(t)) + \int_W \int_K f(t,x(t),u,z) \mu(dz) \lambda(t)(du),\quad t \in T=[0,b], \] \[ x(0) = x_0, \quad\mu \in M_+^1 (K) ,\quad \lambda(t)\in \Gamma(t), \] \[ m_r= \inf_{\lambda \in S_\Gamma} \sup_{\mu \in M_+^1 (K)} \int_0^b \int_K L(t,x(\mu,\lambda)(t),u)\lambda(t)(du)dt. \] Here \(\partial \varphi\) is the subdifferential of the convex function \(\varphi (t, \cdot)\), \(M_+^1(K)\) is the space of probability measures on the set \(K\) modelling the uncertainty in the system produced by the unknown parameter \(z \in K\). The relaxed control \(\lambda (\cdot)\) in (RP) is measurable and \( \lambda \in S_\Gamma \), where \(\Gamma \) = \(\{ \theta \in M_+^1 (W_{\omega}): \theta (U(t)) =1 \}\), \(W_{\omega} = \{ u \in Y : \| u\|\leq \gamma \}\). The sufficient conditions for the equality \( m=m_r \) to be fulfilled (the relaxation theorem) are given. The existence of solutions to the problems (OP) and (RP) are proved.