Estimation of Driver Demand Torque using Parametric and Nonparametric Data-driven Model

Abstract

The complexity of vehicle dynamics is increasing due to growing demands for the inclusion of a large number of functionalities in vehicle model. New generation electronic control units (ECUs) regulate different components of powertrain to produce optimum power and torque necessary to meet the complex functional requirements. But the lookup table or map used in different ECUs e.g., engine ECU, gearbox ECU to create these functionalities are not capable enough to capture the dynamic behavior of system. Thus, effective control of vehicle by ECUs requires a model that is able to accurately predict the dynamic behavior of the system over its complete operating range. The present work proposes both parametric and nonparametric data-driven models that can replace lookup tables or maps used for the estimation of driver torque request. The driver input module estimates the driver demand torque or driver torque request. The inputs to the driver input module are engine speed and accelerator pedal. A data-driven parametric polynomial regression model and nonparametric Volterra model are developed to describe the dynamic behavior of multivariable nonlinear driver input module. The parameters of both the models are estimated using least square optimization algorithm. The input-output data taken from real vehicle dataset is used for both identification and validation of the model. The validation experiments show good fit of the predicted output with actual output. The accuracy obtained from the Volterra and polynomial regression models are 98.27% and 98.6% respectively.