Duality theorem for the stochastic optimal control problem (Q860701)

In generalisation of former papers by several authors but also by T. Mikami and the authors of the present paper, the duality theorem of the stochastic counterpart of the Monge-Kantorovich optimal mass transport problem as well as the properties of the maximisers are studied. To be more precise, for a given non negative continuous function \(L(t,x,u)\) which is convex in \(u\in R^{d}\) and probability measures \(P_{0},P_{1}\) on \(R^{d}\) it is said that the duality theorem for the stochastic counterpart of the Monge-Kantorovich problem \(V(P_{0},P_{1})=\inf \{E[\int_{0}^{1}L(t,X_{t},\beta_{X}(t,X))\, dt]:P_{X_{t}}=P_{t}\), \(t=0,1\); \(X\in \mathcal{A}\}\) (the infimum is taken over all continuous semimartingales \(X\) for which there is some measurable functional \(\beta_{X}\) such that \(W_{t}^{X}=X_{t}-X_{0}-\int_{0}^{t}\beta_{X}(s,X)\, ds\) is an \(\mathbb{F}^{X}\)-Brownian motion) is satisfied if \(V(P_{0},P_{1})=\sup \{\int_{R^{d}}\varphi(1,x) P_{1}(dx) -\int_{R^{d}} \varphi(0,x)P_{0}(dx)\}\) with a supremum that is taken over all classical solutions \(\varphi\) of a certain Hamilton-Jacobi-Bellman equation with \(C_{b}^{\infty}\)-terminal condition \(\varphi(1,.)\). Under appropriate assumptions on \(L\) the authors prove this duality theorem. Applying it the authors then study the properties of the minimiser of the problem \(V(P_{0},P_{1})\). They show in particular that unique minimisers of \(V(P_{0},P_{1})\) can be characterised as the first component of a weak solution of some forward-backward stochastic differential equation. As another application of the duality theorem \(h\)-path processes are considered.