McGill Computer Networks Research Lab

The Call Admission Control (CAC) and Routing Control (RC) problems for communication networks have been formulated as stochastic control problems; due to their complexity, a large number of suboptimal solutions have been proposed for these problems ([1],[2]) and some partial analyses have been carried out. Indeed, because of this complexity, it is impossible to implement even the proposed suboptimal solutions when the entire system model is formulated at any realistic level of detail. However, among the recent analyses of interest, Dziong and Mason [1] obtained suboptimal solutions by assuming statistical independence of the network links, and Marbach et al. [2] obtained an approximation result by the method of neuro-dynamic programming.