Dynamic programming in stochastic control of systems with delay

From MaRDI portal





The authors considers the controlled stochastic differential equation with time delay \[ dX(t)=b(t,X_t,u(t)) dt+ \sigma(t,X_t,u(t)) dW(t), \] where \(X_t=(X(s)\), \(t-\delta\leq s\leq t)\) denotes a segment of the process in the past, \(u(t)\) is an adapted control, \(b\) and \(\sigma\) are suitable Lipschitz mappings and \(W\) is an \(r\)-dimensional Brownian motion. Given the cost functional \[ J(s,\phi,u)=E \left[\int_0^T f(t,X_t,u(t)) dt+h(X_T) \right], \] a dynamic programming principle is proved. For this, a strong and a weak form of the control problem is introduced and strong uniqueness is shown to imply weak uniqueness also in the delay case. As an application, a simple one-dimensional equation is considered with one point delay and one exponentially averaged delay. In this case, an explicit form of the Hamilton-Jacobi-Bellmann function is obtained. The presentation is very accurate and mainly generalizes ideas for stochastic ordinary differential equations.




Cited in
(60)








This page was built for publication: Dynamic programming in stochastic control of systems with delay

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4796610)