Implementation of a Reconﬁguration Algorithm for Cognitive Radio Troy Weingart, Gary V. Yee, Douglas C. Sicker, and Dirk Grunwald Department of Computer Science University of Colorado 430 UCB Boulder, Colorado 80309–0430 Email: Troy.Weingart, Gary.Yee, Douglas.Sicker, Dirk.Grunwald@colorado.edu Abstract— In wireless communication systems, advances in computer architecture and processor technology have made it possible for functionality previously implemented in hardware to become tunable via software. These software-deﬁned radios (SDRs) will allow new radio devices to sense, reason, and adapt to changes in the RF environment and/or application requirements making them cognitive radios (CRs). Fully exploiting the ﬂexibil- ity of cognitive radios, however, requires an understanding of how different permutations of radio parameters impact application- speciﬁc performance metrics. For example, while a CR that is not meeting its bit loss goals could change its operating frequency to reduce the impact of interference. However, the added overhead from changing frequencies could result in an application failing its latency requirements. This paper describes one such method for conﬁguring a cognitive radio and demonstrates the efﬁcacy of the technique on both a simulation based analysis and an in situ evaluation on a software radio platform. Our reconﬁguration system quantiﬁes the inﬂuence of radio parameters such as frequency agility, bit rate, and transmit power for adapting communication at the application, medium access control, and physical layers. The method calls for exhaustively evaluating a set of CR’s conﬁgurations against a variety of performance metrics and applying statistical processes to determine which settings will have the most signiﬁcant impact on performance. Once this is done, the experimental results are then used to inform the design of an algorithm that is able to reconﬁgure to meet performance goals. We show that while the method for deriving the model functions consistently across both the simulation and implementation platforms, the different evaluation platforms emphasize different parameters when controlling the radios. “The views expressed in this article are those of the author and do not reﬂect the ofﬁcial policy or position of the United States Air Force, Department of Defense, or the U.S. Government.” I. I NTRODUCTION Software-deﬁned radios have provided the means to allow mutability in both design and conﬁguration that was once only possible with changes to hardware. This has enabled a broad range of opportunities for improving our use of the radio frequency spectrum. For example, one could have devices that actively reconﬁgure their operation based on applica- tion requirements (e.g., latency or throughput), environmental conditions (e.g., noise ﬂoor) and/or operational policies (e.g., commands to vacate a particular frequency band). Such optimizations, however, require an understanding of parametric cross-layer interactions among the layers within a network protocol stack. This knowledge could then be used to develop a more general purpose reconﬁguration strategy. We have developed a cognitive radio conﬁguration algo- rithm that uses statistical methods to develop a “guided hill climbing” algorithm. The algorithm functions as follows: 1) We start by using a measurement phase to determine the contribution of each system parameter in the cognitive radio to the individual performance metrics (latency, throughput, etc), 2) We use statistical methods based on multivariate analysis to identify the (combined) parameters that most inﬂu- ence performance, 3) We use those most important parameters as input to a control algorithm; the control algorithm varies the radio conﬁguration based on the measured performance metrics, 4) We use stated performance goals to determine if the system is in an acceptable conﬁguration; information about performance metrics are communicated between the radios in the network, 5) Our algorithm attempts to enforce multiple performance goals, such as speciﬁed bandwidth or latency targets, while using the least airtime and power possible By exhaustively evaluating each of the conﬁgurations of a CR against a set of performance metrics, one can apply a statistical process to determine what radio settings had the most signiﬁcant impact on performance. It is important that the resulting performance model be able to account for the inter- actions between parameters and different performance goals. For example, while enabling error correction may improve the experienced bit-loss, it may negatively impact latency beyond a user’s tolerance. We use techniques derived from Design of Experiments (DOE) to determine the potential interaction of input variables on output responses and develop a reconﬁguration algorithm from the produced model [1]. While DOE has historically been applied with great success in improving manufacturing processes, only recently has it been applied in the wireless domain [2], [3]. The technique is applied by ﬁrst identifying