Dynamic decentralized estimation and control in a multi-access broadcast channel

A class of simple recursive procedures for decentralized control of a multi-access broadcast channel is presented. The procedures are designed to maintain a constant level of channel traffic in the presence of backlog fluctuations. The control procedures give rise to simple Markov chain models for the system, and stable throughput is achieved for input rates ?? < e-1. A general technique and some results based on the analysis of drift are introduced for obtaining bounds of geometric type on marginal distributions, hitting time distributions, and occupation measures of random processes. The technique can be used to prove stability of the control policies presented in this paper.