Abstract

Accurate  knowledge  of  vehicle  mass  is  necessary  for improving    control    performance,    driving    stability,    and maintenance  planning.  Since  direct  measurement  requires additional sensors and hardware, virtual sensing has emerged as a practical alternative by inferring mass from signals already available  in  production  vehicles.  In  particular,  longitudinal- dynamics-based methods are attractive because they leverage onboard measurements via the CAN (Controller Area Network) bus. However, existing approaches neglect the variability of gearbox mechanical efficiency and instead treat it as a fixed value during the time-varying operating conditions. Such an assumption  can  deteriorate  mass  estimation  accuracy.  To address this issue, this paper presents a lightweight vehicle mass virtual sensor based on an extended Kalman filter considering mechanical   efficiency   variability.   The   proposed   method incorporates an analytical model of mechanical efficiency so that efficiency changes under different operating conditions can be reflected in real time. By combining this efficiency model with the  EKF  framework,  the  method  effectively  captures  the nonlinear characteristics of longitudinal vehicle dynamics while maintaining  low  computational  complexity.  On-road  driving experimental tests under three different weight conditions are performed to demonstrate the accuracy and robustness of the proposed method. The results show that the proposed method yields more accurate estimates than baseline methods across all tested weight conditions.