Deep Fuzzy System for Dual-Station Target Tracking With Azimuth and Doppler Measurements

Abstract

To address the issues of excessive estimation error and unstable filtering caused by uncertainties in process noise and system models, we propose a Wang–Mendel fuzzy system (WMFS)-based unscented Kalman filter (UKF) algorithm for dual-station target tracking with azimuth and Doppler measurements. The algorithm leverages the strengths of WMFS in handling system uncertainties and complex modeling. During the derivation of the WMFS-UKF, the unscented transform (UT) framework is employed to tackle the nonlinear measurement problem, while the state transition function is reconstructed using a pretrained WMFS. By utilizing the historical states of the target and the expected outputs, the WM fuzzy inference system achieves more accurate state predictions and precise covariance estimates. This leads to significantly improved performance and enhanced stability in target tracking. Simulation experiments and real-data filtering experiment validate the algorithm's effectiveness and robustness in various target tracking scenarios.