An improved fuzzy logic-based vehicle dynamics state estimation framework via coordination of extended Kalman filter and artificial neural networks

This paper proposes a 'combined estimation algorithm (CEA)' designed to estimate the unknown yaw rate and side-slip angle values of a tactical vehicle while navigating at various speeds during a NATO double lane change manoeuvre, using known states. The CEA includes three independent modules: an extended Kalman filter (EKF), artificial neural networks (ANN), and fuzzy logic (FL). The EKF is based on a single-track nonlinear vehicle model, while the ANN using trained data. Both modules take the front axle steering angle as their single input and operate simultaneously in coordination with a novel fuzzy logic (FL) module, which integrates the outputs of both the ANN and EKF, utilising practical rules and membership functions derived from experience and experimental data to enhance prediction performance. Simulations demonstrated that the proposed CEA improves the estimation accuracy by between 8% and 59% for all error metrics, compared to the EKF and ANN alone.