Interference-based scheduling in spatial reuse TDMA

Abstract: Spatial reuse TDMA has been proposed as an access scheme for multi-hop radio networks where real-time service guarantees are important. The idea is to allow several radio terminals to use the same time slot when possible. A time slot can be shared when the radio units are geographically separated such that small interference is obtained. The transmission rights of the different users are described with a schedule.In this thesis we will study various aspects of STDMA scheduling. A common thread in these various aspects is the use of an interference-based network model, as opposed to a traditional graph-based network model. While an interference-based network model is more complex than a graph-based model, it is also much more realistic in describing the wireless medium. An important contribution of this thesis is a comparison of network models where we show that the limited information of a graph model leads to significant loss of throughput as compared to an interference-based model, when performing STDMA scheduling.The first part ot this thesis is a study of assignment strategies for centralized scheduling. Traditionally, transmission rights have been given to nodes or to links, i.e., transmitter/receiver pairs. We compare these two approaches and show that both have undesirable properties in certain cases. Furthermore, we propose a novel assignment strategy, achieving the advantages of both methods.Next we investigate the effect of a limited frame length on STDMA schedules. We first show that the required frame length is larger for link assignment than for node assignment. Further, we propose a novel assignment strategy, the joint node and link assignment, that has as low frame length requirements as node assignment but with the capacity of link assignment.In the last part of this thesis we describe a novel interfence-based distributed STDMA algorithm and investigate its properties, specifically its overhead requirement. In addition we show that this algorithm can generate as good schedules as a centralized algorithm can.