Robust Vehicle Mass Estimation Using Recursive Least M-Squares Algorithm for Intelligent Vehicles

A practical and reliable online vehicle mass estimation scheme is crucial to effectively control an automated and intelligent vehicle. The Recursive Least Squares with Multiple Forgetting Factors (RLSMFF) algorithm has been shown to demonstrate great accuracy and computational efficiency in estimating vehicle mass. Nevertheless, our analysis on the convexity and convergence of the RLSMFF algorithm revealed significant inaccuracies at low sampling rates, as well as an estimation bias when the initial estimate is poor. Thus, the RLSMFF needs high-frequency data and an accurate initial estimate (both which could be challenging to obtain) to generate an accurate estimate. Additionally, the robustness of the RLSMFF algorithm to impulsive disturbances (such as braking, which is a common and inevitable driving maneuver) remains a challenge. To address the aforementioned issues, this paper proposes a robust Recursive Least M-Squares with Multiple Forgetting Factors (RLM-SMFF) algorithm for reliable vehicle mass estimation in the presence of impulsive disturbance. In particular, a restructured longitudinal dynamics model and a bias reduction strategy are introduced to enhance the accuracy of the mass estimation even when the sampling rate is low. An M-estimator is incorporated to suppress the effects of the impulsive disturbance. The stability of our proposed algorithm is also verified mathematically. We evaluated our proposed algorithm using extensive simulations, which showed that our method demonstrates superior accuracy compared to existing mass estimation algorithms, at low computational demand.