Application of Mamdani Fuzzy Inference Systems to Interference Assessments

In dynamic spectrum allocation involving passive wireless systems, such as for weather radiometry or radio astronomy, estimating potential interference is crucial in setting transmission spectral and spatial limitations for potentially interfering transmitters. The challenge in accurately assessing interference is heightened by variability in environmental factors and limitations of static modeling. This often leads to protection levels that are either excessively stringent or overly permissive. Dynamic spectrum access (DSA) systems commonly rely on the ability to precisely model transmissions and estimate interference prior to frequency assignment, where total interfering power is acquired by means of summing individual contributions to a potential victim receiver. As an alternative to the worst-case static calculations, this paper proposes the implementation of a Mamdani-type fuzzy inference system as the assessment mechanism for interference levels. In this approach, transmitter operations and network characteristics are characterized by their degree of membership with various linguistic variables. Membership grades are then provided to a ruleset determined by the expected relationship between input and output parameters. The value implied by the rules gives an estimation of interfering power level that may be tuned by adjusting the membership characterization of parameters. After tuning, simulation results yield a Root Mean Square Error (RMSE) improvement of approximately $39 \%$, demonstrating the system’s ability to adapt to varying levels of agreement with static calculations.