A numerical approach to the infinite horizon problem of deterministic control theory (Q752463)

The author considers the closed loop control of the infinite horizon problem of deterministic control theory. A time discretization of the related Hamilton-Jacobi (HJ) equation was introduced by \textit{I. C. Dolcetta} [ibid. 10, 367-377 (1983; Zbl 0582.49019)]. At this stage, the author introduces a discretization of the state variable using finite element techniques. The convergence of the whole approximation process to the viscosity solution of the HJ equation is demonstrated. A relaxation type algorithm proposed by \textit{R. L. Gonzalez} and \textit{E. Rofman} [SIAM J. Control Optimization 23, 242-266, 267-285 (1985; Zbl 0563.49024, Zbl 0563.49025, resp.)] is used to obtain an approximate solution of the HJ equation. It is shown that it can be reinterpreted as an acceleration method for a sequence generated by a contracting operator. This allows an error estimate for each step of the algorithm to be obtained.