Stochastic drift counteraction optimal control and enhancing convergence of value iteration
Robert A. E. Zidek, Ilya V. Kolmanovsky · 2016
This paper considers stochastic optimal control of a class of nonlinear discrete-time systems with disturbances modeled by a Markov chain. The objective is to maximize the expected time or the expected total yield until prescribed constraints are violated. Conditions for the existence of an optimal solution are derived, and a new algorithm is developed that converges to the optimal solution faster than conventional value iteration / dynamic programming. Two numerical examples of stochastic adaptive cruise control and glider flight management are treated.