Efficient Generation of Spatially and Frequency Correlated Random Values for Cognitive Radio Network Simulators

We address the generation of random values exhibiting both frequency-dependent spatial correlations and position or more generally terminal-dependent frequency correlations. This case is relevant for the development of radio system/network simulators operating over many bands or over a large spectrum, such as cognitive-radio-inspired networks. The simplistic Kronecker approximation does not allow accounting for both dependencies, but it is shown that a method that is nearly as simple (semi-Kronecker approximation) improves the quality of the generation process very much. The method makes use of nonassociative left and right products of user-dependent frequency correlation matrices and of frequency-dependent spatial correlation matrices. To further reduce the computational load, a correlation zone growth algorithm is also developed, which allows limiting the size of the geographical area within which the correlations are computed. These two features provide high efficiency in the generation process of frequency- and user-domain correlated random values.