Development and analysis of a probabilistic reasoning methodology for spectrum situational awareness and parameter estimation in uncertain environments

Advancements and proliferation of wireless devices and capabilities have expanded the need for spectrum situational awareness in support of mobile applications. Accurate representations of spectrum usage parameters, however, are limited by the ability to attain sufficiently accurate information in environments characterized by significant uncertainty. This paper describes the design and characterization of a probabilistic reasoning methodology for spectrum situational assessment. The approach uses Functional Causal Models — a form of Bayesian Networks — to represent the propagation environment and enables parameter estimation in uncertain environments. The general model is described and a simulation implementation is used to as a basis for quantitative and qualitative characterization. Results demonstrate the degree of uncertainty reduction for various parameters as functions of prior beliefs and consistency with theoretical predictions. Path loss estimation error was significant reduced to within 10 dB of true conditions via Bayesian updating with several cases showing errors of less than 5 dB. Estimations errors of transmitted power were marginally reduced in the selected scenarios, and path distance could not be reliably estimated. Thus the proposed approach produces path loss estimates that could enable applications such as Dynamic Spectrum Access systems to operate with acceptable levels of risk.