Adaptive Coefficient Size-Based Generalized Loss Extended Kalman Filter

While current self-tuning robust Kalman estimators successfully manage uncertain noise parameters and anomalous measurements, their precision declines significantly when processing intricate non-Gaussian distributions with heavy-tailed characteristics. Advancing beyond fixed-form robust filters, this research develops an adaptively scaled extended Kalman filter (AGLEKF) incorporating a reconfigurable generalized loss mechanism. The GL’s distinctive feature is its dynamic functional reshaping capability through parametric adaptation, providing superior handling of complex outlier scenarios compared to conventional static loss structures. Experimental validation through GPS/INS integrated navigation demonstrates that the AGLEKF filter achieves higher accuracy than existing filters in complex noise environments.