Adaptive Estimation of Signals of Opportunity

To exploit unknown ambient radio frequency signals of opportunity (SOPs) for positioning and navigation, one must estimate their states along with a set of parameters that characterize the stability of their oscillators. SOPs can be modeled as stochastic dynamical systems driven by process noise. The statistics of such process noise is typically unknown to the receiver wanting to exploit the SOPs for positioning and navigation. Incorrect statistical models jeopardize the estimation optimality and may cause filter divergence. This necessitates the development of adaptive filters, which provide a significant improvement over fixed filters through the filter learning process. This paper develops two such adaptive filters: an innovation-based maximum likelihood filter and an interacting multiple model filter and compares their performance and complexity. Numerical and experimental results are presented demonstrating the superiority of these filters over fixed, mismatched filters.