This paper describes a novel particle filter technique to geolocate radio frequency emitters with unknown emitter frequency using passive, Doppler-shifted frequency measurements. Doppler-based geolocation algorithms suffer in general, from the non-linear and coupled relationship between the unknown emitter location, unknown emitter frequency, and observed Doppler-shifted frequency measurements. This non-linearity precludes a computationally efficient, closed-form implementation and directs algorithmic solutions toward robust but inaccurate and computationally expensive grid-based techniques. The new method leverages the rapid particle convergence properties developed in the simultaneous location and mapping (SLAM) community with the robustness of grid-based solutions to define a new more effective importance density. The new method's performance is compared against a standard sampling importance resample (SIR) particle filtering implementation and the Cramer-Rao Lower Bound. Results indicate that the new approach can significantly improve the convergence rate with far fewer particles.
展开▼
