Numerical computation of the optimal feedback law for nonlinear infinite time horizon control problems (Q1570399): Difference between revisions

This article presents an approach for the numerical computation of optimal feedback laws for nonlinear infinite horizon optimal control problems of the type \[ \text{ minimize} \quad J(x_0,u)=\int_0^\infty \left(h(x(t,x_0,u)) + u(t)^Tu(t) \right)\,dt, \] where \(x(t,x_0,u)\) is the solution trajectory of a nonlinear affine control system. The approach proceeds in two steps: First, the canonical (state and costate) equations for the Hamiltonian related to this optimal control problem are solved by a multiple shooting method for several mesh points \(x_0\). Second, the results of the first step are used to obtain an approximately optimal control law at the mesh points which then is interpolated in order to compute the approximating optimal feedback law. After several details regarding the implementation of these two steps are discussed, the author turns to the description of extensive numerical tests, which show the application of the algorithm to a number of control problems. While these numerical tests show good results, a formal convergence proof and an estimate for the discretization error of this method seem to be difficult, as the author himself points out in the concluding remarks.