Approximate policy iteration for semiconductor fab-level decision making - a case study

In this paper, we propose an approximate policy iteration (API) algorithm for a semiconduc-tor fab-level decision making problem. This problem is formulated as a discounted cost Markov Decision Process (MDP), and we have applied exact policy iteration to solve a simple example in prior work [1]. However, the overwhelming computational requirements of exact policy iter-ation prevent its application for larger problems. Approximate policy iteration overcomes this obstacle by approximating the cost-to-go using function approximation. Numerical simulation on the same example shows that the proposed API algorithm leads to a policy with cost close to that of the optimal policy