Receding horizon directional unscented filter for heavy-duty vehicles incorporating sensor modeling constraints

Abstract

This paper proposes a new wheelbase and mass estimation algorithm called the receding horizon-based directional unscented filter (RHDUF) algorithm, which is developed based on the lateral dynamics model. The unscented Kalman filter (UKF), which is widely used for estimating nonlinear systems, has an infinite impulse response structure. However, the UKF is vulnerable to uncertainty and accumulates errors gradually. To address this problem, filters with a receding horizon structure can be introduced, but they may lead to a heavy computational burden, which is not desirable for fast estimation. To overcome this problem, we propose a new sigma-point distribution method combined with the receding horizon structure in this paper. The formula is derived through the measured sensor value and the modeling equation of the sensor, and the direction of the sigma points is determined using this equation. This formula can improve performance and reduce the increase in the computation time caused by the receding horizon structure by reducing the number of sigma points simultaneously. The accuracy of the new estimation algorithm is verified through an experiment for heavy-duty vehicles.