The idempotent analog of resolvent kernels for a deterministic optimal control problem (Q5951101)

It is well known that the discrete Bellman equation can be treated as linear over appropriate idempotent semirings, and in this paper, one of the possible directions arising from idempotent analysis is explored: namely, a solution of a discrete Hamilton-Jacobi-Bellman equation is represented in terms of idempotent analysis as a convergent series of integral operators. The main result obtained is the representation formula for the value function in terms of linear operators. This permits to show important properties and to consider the issue of using parallel algorithms developed for linear algebra in nonlinear cases. A dynamic process where we can act by choosing control functions and stopping times is considered. The Hamilton-Jacobi equation associated with the stopping time problem is a variational inequality of obstacle type. After the discretization of the optimal control problem, using an appropriate algebraic structure for linearizing, the problem is rewritten and the analog of the Dirichlet formula for the idempotent calculus is found. The proofs of the main results are based on the integral representation of the solution.