Fluid flow model and queueing analysis of an error control scheme for wireless broadband access networks

Abstract

The subject of this paper is the development of a traffic model for indoor wireless broadband access networks, which can be considered as the wireless extension of a fixed quality-of-service oriented network (e.g., based on the ATM asynchronous transfer mode protocol). Mobile users communicate with corresponding users in the fixed network via an access point. A central scheduler in the base station is responsible for the fair assignment of transmission capacity to the users according to a dynamic time division multiple access scheme. In order to improve the high and time-variant probability of transmission errors, an error control scheme based on an efficient combination of a go-back-N and a selective repeat ARQ (automatic repeat request) scheme is applied. Additionally, the protocol features a flow control mechanism which dynamically adapts the transmission rate of a connection to the link quality. This prevents the channel from being blocked by a user with momentarily high transmission error probability, and thus considerably increases the total system throughput. The traffic model developed in this paper is based on the fluid flow approximation and provides a stochastic process describing the arrival stream of a connection taking into account both the bursty characteristics of the traffic sources and the retransmission traffic. The retransmission traffic is modelled by a realistic radio channel, taking into account the time-variant fluctuations resulting from the movements of the user. The traffic model allows one to calculate the total system load, including the re-transmission traffic. It turns out that there is good agreement between the calculated values and those obtained by simulation. Furthermore, the traffic model is used for a queueing model of the wireless broadband access network based on the diffusion approximation. The queueing analysis provides a good approximation of the distribution of the buffer length observed in the real system. Copyright © 2003 AEI.