Performance evaluation of cognitive radio (CR) networks is an important problem but has received relatively limited attention from the CR community. Unlike traditional radios, a cognitive radio may change its objectives as radio scenarios vary. Because of the dynamic pairing of objectives and contexts, it is imperative for cognitive radio network designers to have a firm understanding of the interrelationships among goals, performance metrics, utility functions, link/network performance, and operating environments. In this paper, we first overview various performance metrics at the node, network, and application levels. From a game-theoretic viewpoint, we then show that the performance evaluation of cognitive radio networks exhibits the interdependent nature of actions, goals, decisions, observations, and context. We discuss the interrelationships among metrics, utility functions, cognitive engine algorithms, and achieved performance, as well as various testing scenarios. We propose the radio environment map-based scenario-driven testing (REM-SDT) for thorough performance evaluation of cognitive radios. An IEEE 802.22 WRAN cognitive engine testbed is presented to provide further insights into this important problem area.

This paper proposes and evaluates a new method for mitigating inter-cell interference. We show that the proposed method results in a minor degradation of block error rate under low traffic loading but results in significant improvements for high traffic loads. We also propose and compare two options for resource re-allocation following allocation conflicts: immediate (synchronous) re-allocation and round-robin re-allocation. Simulation results indicates that round-robin reallocation results in good BLER performance when the traffic load is high.

We show that the fixed power, synchronous interference avoidance (IA) scheme of (C. Rose et al, IEEE Trans. on Wireless Comm., vol.1, no.3, p. 415-427, 2002) employing the (greedy) eigen-iteration can be modeled as the recently developed potential game of (D. Monderer et al, Journal of Games and Economic Behavior, vol.14, no.0044, p.124-143, 1996). Motivated by the fact that receivers can make small mistakes, we consider the convergence of the eigen-iteration when noise is added in a manner similar to (P. Anigstein, IEEE Trans. On Inf. Theory vol.49, no.4, 2003). Further, we restrict ourselves to a class of signal environments that we call levelable environments. Applying game-theory, we obtain a convergence result similar to that of the Anigstein method, for levelable environments: arbitrarily small noise assures that the eigen-iteration almost surely converges to a neighborhood of the optimum signature set.

In this paper, we examine the conditions and behavior of several common convergence dynamics from game theory and show how they influence the structure of networks of cognitive radios. We then apply these to previously proposed distributed power control algorithms and describe how they impact network complexity.