Reference Velocity Estimation with Variable Gain Based on Powertrain Dynamics for Production Hybrid Electric Vehicle

Abstract

<div class="section abstract"><div class="htmlview paragraph">Reference velocity (i.e. the absolute velocity of vehicle center of gravity) is a key parameter for vehicle stability control functions as well as for the powertrain control functions of hybrid electric vehicle (HEV). Most reference velocity estimation methods employ the vehicle kinematic and tire dynamic equations to construct high order linear or nonlinear model with a set of parameters and sensor measurements. When using those models, delicate algorithm should be designed to prevent the estimates from deviating along with the increase of nonlinearity, modeling error and noise that introduced by high order, parameter approximation, and sensor measurements, respectively. Alternatively, to improve the function robustness and calibration convenience, a straightforward online estimation method is developed in the paper by using a second-order powertrain dynamic model that only need a small set of vehicle parameters and sensor values. First, the HEV powertrain dynamic model is established for the vehicle longitudinal velocity estimation. Second, a classic Luenberger observer with variable estimation gains are designed. Third, the variable estimation gains are scheduled based on the vehicular operational conditions to determine whether the estimates need to be dominated by the dynamic model or by the measurements in different condition. Then the algorithm is integrated into the vehicle control unit (VCU) of a mass production HEV, which is a powertrain supervisory controller that possesses all the control inputs and measurements signals needed by the observer. Finally, the estimation accuracy is verified by experiments on both high- and low-μ (-adhesion) road, such as the snow surface, ice surface, and urban concrete pavement, etc. Due to the low order and minor parameters and measurements needed, as well as the variable estimation gain scheduled with operational conditions, the algorithm robustness and calibration convenience are guaranteed.</div></div>