Discontinuous Solutions of the Hamilton--Jacobi Equation for Exit Time Problems (Q4507397)

Continuing previous work in [\textit{J. J. Ye} and \textit{Q. J. Zhu}, Nonlinear Anal., Theory Methods Appl. 34, No. 7, 1029-1053 (1998; Zbl 0960.49022)], the author proves several new results concerning the value function, \(V(.)\), of the so-called ``exit time problem'' which consists in the minimization of the functional: \[ J(x,u):=\int_0^{t^*}e^{-\lambda s}f(y(s),u(s)) ds +e^{-\lambda t^*}h(y(t^*)) \] subject to: \[ y'(t)=g(y(t),u(t)), \;u(t)\in U \text{ a.e. }([0,t^*]), \quad y(0)=x\in E, \] \[ y(t)\in E\subset R^n \;\forall t\in [0,t^*), \quad y(t^*)\in E^c:=R^n\setminus E. \] Instead of using the ``Fréchet semidifferentials'' as in the well-known theory of viscosity solutions [e.g., \textit{M. Bardi} and \textit{I. Capuzzo-Dolcetta}, ``Optimal control and viscosity solutions of Hamilton-Jacobi-Bellman equations'' (1997; Zbl 0890.49011)], the author uses the relatively recent concept of ``proximal semidifferential'' in [F. H. Clarke, ``Methods of dynamic optimization'' (1989; Zbl 0696.49003); \textit{P. D. Loewen}, ``Optimal control via nonsmooth analysis'' (1993; Zbl 0874.49002)] to introduce ``proximal generalized solutions'' for the associated Hamilton-Jacobi-Bellman equation: \[ \lambda V(x)+H(x,-\nabla V(x))=0, \;x\in E, \quad V(x)=h(x), \;x\in \partial E,\tag{HJB} \] defined by the Hamiltonian \(H(x,p):=\max_{u\in U} \{pg(x,u)-f(x,u)\}.\) The first main result is Theorem 2.2 stating that under fairly general hypotheses on the data, including the case \(\lambda =0\), the ``l.s.c. envelope'' \(V_*(x):=\liminf_{y\to x}V(y)\) is a ``proximal supersolution'' while the ``u.s.c. envelope'' \(V_*(x):=\limsup_{y\to x}V(y)\) is a ``proximal subsolution'' of (HJB), each of them satisfying suitable mixed boundary conditions expressed in terms of proximal semidifferentials. Some refinements of this result are obtained in Theorems 2.3, 2.4 assuming that the value function itself is either l.s.c. or u.s.c. while in Theorem 2.7 a uniqueness result for l.s.c. proximal solutions is proved under an additional hypothesis on the data.