Parametric optimization of software defined radio configurations using design of experiments

Cognitive radio (CR) systems incorporate learning and decision making into wireless and networking systems with the goal of improving performance and interoperability. Research has focused on artificial intelligence control and optimization of radio input parameters with little attention placed on identifying initialization parameters of cognitive engines or on testing methods. While CR techniques continue to advance, calibration and testing remain largely stagnant with reliance on ad hoc and highly application specific approaches. Given that cognitive radio will be deployed in a variety of environments with each requiring unique calibration, systematic procedures are needed. An approach founded in design of experiments provides a purposeful framework for performing testing and identification of initialization parameters with an efficient number of test cases. Response surface methodology designs identify representative knowledge of system performance including input parameter significance and quadratic estimation models of output metrics. An example of calibrating transmit-and-receive gain settings on a software-defined radio illustrates the use of the framework.