A transformation approach for solving the Hamilton-Jacobi-Bellman equation in \({\mathcal H}_2\) deterministic and stochastic optimal control of affine nonlinear systems. (Q1398404)

The author studies the controlled stochastic equation \[ dX_t= \{f(t, X_t)+ g(t, X_t)u_t\}\,dt+ h(t, X_t)\,dW_t \] (with smooth coefficients \(f\), \(g\), \(h\), a driving Brownian motion \(W\) and control process \(u\) taking its values in a compact subset of \(R^k\)) endowed with a quadratic cost function. He presents a transformation approach for solving the associated Hamilton-Jacobi-Bellman equation. The key idea consists in the reduction of the equation into a set of coupled algebraic-differential inequalities, using a suitable parametrization. If they are solvable, a solution of the Hamilton-Jacobi-Bellman equation can be constructed. The idea has been developed by the author in an earlier paper, and has first been applied in the deterministic counterpart of the above problem, the deterministic \({\mathcal H}_2\) control problem and its associated Hamilton-Jacobi-Bellman equation; an overview about this is given in the first part of the paper.