Throughput Analysis of Interfering Packet Radio Networks

Abstract: In this thesis, a probabilistic framework is presented and used for analysis of the throughput of interfering packet-based radio networks, referred to as packet radio networks (PRNs). Previous research on the performance of PRNs covers a wide variety of systems and aspects in varying levels of detail. However, the work presented in the literature has mostly been limited to fixed packet lengths. The framework presented in this thesis allows for detailed analysis of PRNs using different lengths of the transmitted packets, which is the typical situation in real systems. The framework also allows for analysis of PRNs using slow frequency hopping and different spectral widths and shapes of the transmissions, which enable analysis of heterogeneous systems of interfering networks. An introduction to radio systems and interference is presented along with an overview over previous work on interfering PRNs. An overview of the analytical framework used in this thesis is also presented, including a description of the system model and two methods used for deriving closed form throughput expressions. The first method is based on packet collisions, where two or more packets overlapping in time and frequency result in lost packets and reduced throughput. The second method is based on the amounts of interfering energy received by the network units. Here, a packet is assumed to be lost if the received amount of interfering energy exceeds a given threshold. Assuming that packet collisions always result in lost packets, exact and approximative closed form expressions are derived for both successful packet reception probabilities and throughput of interfering networks. The closed form expressions are used in analyses of interfering slow frequency-hopping networks, illustrating how the relative lengths of packet transmissions can have a great impact on the achievable data rates. Closed form expressions for the throughput of interfering networks based on received interfering energy quantities are also derived. These can be used to analyze heterogeneous systems of interfering networks where different radio interfaces are used by each network, a situation which is becoming increasingly common in shared frequency bands. Finally, the relation between the two methods of analysis is investigated, and in addition, results from calculations are compared with results from measurements on real networks.