Abstract
In this paper it is shown that a modified Gauss-Seidel-Algorithm with exclusion of suboptimai actions can be used for approximative solving a continuously discounted Semi-Markovian decision problem. Moreover it is proved that the overrelaxation factor of the algorithm introduced in [5] can be improved. If the sojourn times are constant and the overrelaxation factor is independent of states and actions one obtains results of Reetz [6,7] for (discrete) discounted Markovian decision problems. Finally, an example illustrates this method including exclusion of suboptimai actions
*Corresponding Author
*Corresponding Author
Notes
*Corresponding Author