Resource-Efficient Relay Selection in Cooperative Wireless Networks

This thesis focuses on cooperative relaying protocols employing a relay node that overhears data transmissions between a source and a destination nodes and retransmits the data when necessary. The achieved signal diversity at the destination improves the data recovery in fading-rich environments. While cooperative relay can improve communication on one link, the induced relay interference can decrease the overall network capacity. In Chapter 3 of this thesis a contention-based relay selection method is proposed that assigns a relay which retransmissions have low impact on neighboring nodes. By using one of the proposed contention and selection functions, a higher spatial channel reuse in uniform and clustered networks is achieved. The main role of relay selection is to timely provide a relay that maximizes certain performance metrics. However, the required coordination overhead can diminish performance benefits anticipated from cooperation. Chapter 4 presents an analytical framework for modeling cooperative relaying with relay selection using semi-Markov processes. The comparison of four relay update schemes shows that the required selection overhead can significantly decrease throughput and energy efficiency of cooperative relaying when selections are performed frequently and overhead is large. A proposed adaptive selection scheme, which triggers a new selection only when the cooperative link fails, is shown to have the lowest selection rate among the compared schemes in slow-fading channels. Chapter 5 provides an experimental study of cooperative relaying in industrial wireless sensor networks. Cooperative relaying with three relay selection schemes is implemented in off-the-shelf IEEE 802.15.4 devices. Measurements show that all cooperative protocols outperform non-cooperative transmissions in terms of delivery ratio and delay. A trace-based analysis is used to demonstrate how system parameters can be adjusted to improve the delivery ratio and reduce the number of triggered selections.