Abstract

This paper presents a novel adaptive approach to the classical Kalman filter algorithm, aimed at improving estimation accuracy in dynamic systems subject to varying noise characteristics. The proposed method, termed Error-Adaptive Kalman Filter (EAKF), dynamically adjusts the process noise covariance matrix based on real-time innovation sequence analysis. Our approach addresses a common limitation of standard Kalman filters: their reliance on fixed noise parameters that may not accurately represent time-varying system conditions. Simulation results demonstrate that the EAKF achieves a 7.2% improvement in position estimation and 3.1% in velocity estimation, along with a more substantial 20.8% improvement in estimation consistency compared to the standard Kalman filter when applied to nonlinear trajectory tracking problems. The computational overhead is negligible at around 5%, making our method suitable for real-time applications in navigation, tracking, and control systems.