METHOD FOR OPTIMIZING COMPLEX SIGNAL ENSEMBLES IN COGNITIVE NETWORKS

Abstract

The paper proposes a method for optimizing ensembles of complex signals in the time domain based on the integration of stochastic search with local nonlinear optimization. The method is aimed at reducing mutual signal correlation and equalizing the energy–spectral characteristics of the ensemble under conditions of stochastic uncertainty and interference in cognitive telecommunication networks. Unlike existing approaches that consider only linear models or static parameters, the proposed method implements a multistage adaptation of signal time intervals using stochastic search for global exploration of the solution space and local optimization (gradient descent and Levenberg–Marquardt algorithm) for fine-tuning parameters within the identified subregions.

Analytical dependencies are developed for evaluating cross-correlation and energy indicators of the ensemble, ensuring a balance between interference immunity and ensemble volume. The signal permutation mechanism between time intervals has been improved through the introduction of a modified superfactorial that incorporates weighting coefficients of intervals and correlation constraints, allowing minimization of mutual interference without reducing structural diversity.

Experimental modeling on LTE and 5G NR signals demonstrated that increasing the number of time segments from 0 to 16 reduces the average mutual correlation coefficient from 0.42 to 0.08, which corresponds to quasi-orthogonality conditions (ε ≈ 0.1). The use of local optimization improved ensemble stability under varying environmental parameters, while the modified permutation procedure reduced correlation peaks by up to 27%.

Thus, the proposed method provides comprehensive optimization of complex signal ensembles by combining the advantages of stochastic and deterministic approaches, making it suitable for application in adaptive telecommunication environments with multiple access and dynamic channels.