On a system of Hamilton-Jacobi-Bellman inequalities associated to a minimax problem with additive final cost (Q1774379)

The authors study in the framework of viscosity solutions the ``minimax optimal control problem'' which consists in the {minimization} of the cost functionals \[ J(t,x,\alpha(.)):=\Psi(y(T))+\text{ess}\sup_{s\in[t,T]}f(s,y(s),\alpha(s)), \;(t,x)\in [0,T]\times \mathbb R^m \] subject to: \[ y'(s)=g(s,y(s),\alpha(s)), \;\alpha(s)\in A\subset \mathbb R^q, \text{ a.e. }([t,T]), \;y(t)=x. \] As in some previous papers, in order to use the dynamic programming method and, consequently, the theory of viscosity solutions, the authors extend the state-space by introducing an auxiliary state variable: \[ y_{m+1,\alpha(.)}(\tau,t,x,\rho):=\begin{cases} \rho &\text{if } \tau=t \\ \max\{\rho,\text{ess}\sup_{s\in [t,\tau]}f(s,y(s),\alpha(s))\} & \text{if } \tau\in (t,T] \end{cases} \] which defines a new (``extended'') optimal control problem, of Mayer type, whose value function is defined by: \[ v(t,x,\rho):=\inf_{\alpha(.)}[\Psi(y_{\alpha(.)}(T,t,x))+ y_{m+1,\alpha(.)}(T,t,x,\rho)] \] and which is proved to be the unique viscosity solution of a rather complicated system of Hamilton-Jacobi-Bellman inequalities; it is proved also that if \[ m_f(t,x):=\min_{a\in A}f(t,x,a) \;\forall \;(t,x)\in [0,T]\times \mathbb R^m \] then the value function, \(u(.,.)\), of the original problem is given by: \[ u(t,x)=v(t,x,m_f(t,x)) \;\forall \;(t,x)\in [0,T]\times \mathbb R^m. \]