Wireless ATM (W-ATM) has recently gained much attention due to the increasing demand for the support of different services in future personal communication network (PCN). As an extension of the wired ATM network, W-ATM should support the features that are provided in standard ATM networks such as supporting of broad range of services and QoS requirement for different services. However, it makes the guarantee of QoS requirement more difficult in wireless environment because of limited bandwidth and fluctuating noise level. To prevent the waste of bandwidth caused by sending ATM cell in the time of bad channel quality and reduce the delay for real-time service due to retransmission, an effective scheduling algorithm in W-ATM should be able to estimate the channel quality and allocate the...[ Read more ]

Wireless ATM (W-ATM) has recently gained much attention due to the increasing demand for the support of different services in future personal communication network (PCN). As an extension of the wired ATM network, W-ATM should support the features that are provided in standard ATM networks such as supporting of broad range of services and QoS requirement for different services. However, it makes the guarantee of QoS requirement more difficult in wireless environment because of limited bandwidth and fluctuating noise level. To prevent the waste of bandwidth caused by sending ATM cell in the time of bad channel quality and reduce the delay for real-time service due to retransmission, an effective scheduling algorithm in W-ATM should be able to estimate the channel quality and allocate the bandwidth according to the service characteristics and channel quality.

The objective of this thesis is to design an adaptive scheduling algorithm for W-ATM networks. The schemes that are proposed in this thesis are the extension of the two-state adaptive scheduling algorithm, which is proposed by Hui and Tsang. According to the two-state adaptive scheduling algorithm, the assigned fair-queuing weights are changed adaptively according to the service type and the extra bandwidth needed for error protection in the current channel condition. To further improve that scheme, three different multi-state adaptive scheduling algorithms are proposed and have been shown to produce good performance in Rayleigh fading channel.

An analytical model for the multi-state adaptive scheduling is also developed. The system is modeled by a two dimensional Markov chain. Numerical results show that the model can accurately estimate the cell loss rate and throughput of CBR traffic.