A cognitive engine (CE) is an intelligent package that turns the knobs and reads the meters of a controllable radio system. As discussed in Chapter 1, this concept was initially introduced as a software entity that interacts with an electronically configurable radio transceiver [1]. Early examples exhibited limited functionality and software portability, often focusing on fairly specific problems or approaches that were very dependent on particular aspects of the support platform. Subsequent work on the topic has developed many of the trends pioneered by these early examples and expanded the domain of a CE. However, it is important to consider that not much more than a decade has passed since the inception of the concept of a CE for radio application; there are a great many challenges left unsolved and capabilities left undiscovered. In fact, the full scope of CE is only beginning to develop and there is a vast territory left to explore.

In selecting computational hardware for SDR/CR applications, there is not necessarily a single platform and/or combination of computational devices that address an application. Having a particular platform that is readily available, familiarity with a particular set of computational devices, cost, time-to-market, etc. will influence a developer's choice of a platform. By understanding the underlying computational device architecture and the associated tradeoffs designers can better understand how to leverage and utilize the devices in their systems. The use of MoCs ultimately provides an appropriate framework to analyze, develop, and integrate application development. MoCs provide various design capabilities, guarantees, and programming complexity. In some applications, a combination of computational devices might be necessary, and a combination of computational models might also be necessary such as when bridging the reactive and real-time system perspective between CEs and SDRs.

Herein we have tangled with the challenging topic of performance evaluation as it relates to CRs. The primary feature of this process is the estimation of the performance function which is inherent to the operation of a CR and determines the reactions of the CR to any given scenario. Through estimation of this function, an evaluator determines the information necessary to make ajudgment about the goodness of the CR's operation. We have discussed the major factors and metrics that influence this function and provided a framework for their consideration.

In this chapter, we focus on representative cognitive radio applications where working prototype hardware and software exist or where standards or regulations are available to guide the development of such prototypes. We have not attempted to cover the many proposed applications that have been simulated but not yet adopted or built or for which standards work is still in its early stages. Our examples cover the top level of the application-driven taxonomy proposed in Reference 1: dynamic spectrum access, wireless network management, user-aware wireless services, and homogeneous networking. See Reference 2 for an alternative treatment of many of the same applications.