Signal processing for application-specific ad hoc networks

[ The role of signal processing in protocol design ] T he layered architecture used in large data communication networks provides modularity at the expense of efficiency. This paradigm has been successful in wired and many wireless networks. Modularity simplifies the design of large heterogenous and complex networks via a " divide-and-conquer " approach, decoupling an enormously difficult problem into tasks that can be pursued independently. Because of this, new applications can be developed independently of the medium access and routing strategies specified at lower layers, and new physical layer techniques can be implemented without changing upper layer implementations. Perhaps the most remarkable feature of the layered architecture is that it makes the network design scalable. A prime example is the Internet, which has grown from a hand full of nodes in the ARPAnet [3] to hundreds of millions of nodes today. But there is a different kind of network, one that is designed for specific applications. In contrast to the design of the Internet, serving individual nodes is not always the ultimate objective. Consider, for example, a sensor network deployed for target detection and tracking, environmental monitoring, or the detection of a specific chemical compound. In these applications , network performance should not be measured by general purpose metrics such as the data rate at the link level or by the throughput over the network. Conventional performance metrics such as throughput and delay do not necessarily translate to a performance measure suitable for signal detection and estimation applications. For application-specific networks, and sensor networks in particular, performance should be measured instead by application-defined metrics such as the miss detection and false alarm rates, the network lifetime for performing these tasks, and the energy efficiency of target detection, tracking , and estimation. If an application-specific metric is to be optimized, signal processing may have a role in defining network architectures