MAC and Routing in Wireless Ad Hoc and Sensor Networks: a Cross-Layer Approach

"Respecting a network architecture yields better guarantees of reliability, longevity, and modularity, but much better performance can be potentially achieved through wisely chosen violations to that architecture." In a nutshell, this is the message of a recent paper (see [1] in Chapter 1) outlining pros, cons, consequences and risks of cross-layer design, a currently widely adopted paradigm for wireless networks. The increasing attention and momentum that cross-layer design has recently gained is explained by its potential advantages, namely the network performance improvements that can be achieved, especially under stringent constraints in terms of hardware and computational power. A short definition of cross-layer design identifies this technique as a means of performing information exchange among different layers in the classic ISO/OSI protocol stack model, and of harvesting the potential design opportunities and performance improvements that follow. However, by breaking the modular structure of the ISO/OSI stack, one may encounter two orders of problems: first, unwanted interactions may be introduced; second, the generality of the architecture is lost. While a careful design phase can overcome the first problem, the second one requires stronger efforts. In fact, any cross-layer design is inherently specific to the type of network and scenario it is applied to, and limits the performance improvements to that specific type. Due to this loss of generality, the same protocol hardly offers the same results as applied to different types of networks. In this Thesis, we will show two relevant examples of successful cross-layer design applied to two very different kinds of wireless networks. The first example deals with ad hoc networks with multiple antennas and MIMO communications. Due to the specific scenario, it can be assumed that nodes have high throughput needs and can accept to, e.g., spend more energy in performing the processing required by MIMO signaling in order to achieve greater communication speed. The analysis of this scenario is focused on the design of a novel PHY-aware MAC protocol for MIMO ad hoc networks and on the analysis and optimization of its performance. A completely different point of view is required instead to handle wireless sensor networks (WSNs), the second type of wireless network considered in this Thesis. Peculiar to WSNs are the usually low communication speed, processing capabilities and energy supplies. Among others, these constraints do not allow complicated signal processing or the storage of a large amount of information. In turn this requires to limit the buffer of the nodes (the sensors hence have only a limited packet queue) and also to design protocols whose "state" can be summarized and efficiently held in the limited memory of the sensors. In the Thesis, we will provide an in-depth analysis of a geographic MAC/routing protocol for WSNs, and build upon it to yield a complete solution for channel access and packet forwarding. Part of this study is the design of an algorithm to route packets around connectivity holes, where geographic protocols alone fail. In the appendix, the same cross-layer design concepts are applied to wireless underwater networks, a particular instance of WSNs where communications take place over long delay, low rate acoustic channels, and incur strongly frequency-dependent channel effects. All results (analysis, simulations, comparisons with other solutions) show that cross-layer design is in fact very effective, and offers valuable opportunities to leverage specific features that can lead to performance improvements in each kind of wireless network.