On optimization of polling policy represented by neural network

Abstract

This paper deals with the problem of scheduling a server in a polling system with multiple queues and complete information. We represent the polling policy by a neural network; namely, given the number of waiting customers in each queue, the server determines next queue he should visit according to the output of the neural network. By using the simulated annealing method, we improve the neural polling policy in such a way that the mean delay of customers is minimized. Numerical results show that the present approach is especially valid for asymmetric polling systems whose analytical optimization is considered intractable.