Genetic algorithm parameters estimation applied to vehicle powertrain dynamics

Typically, in the automotive field, classic estimation or filtering techniques (e.g. Least Squares, Kalman Filter) are used to characterize models. However, in scenarios in which there is a lack of data or some quantities are not available from CAN (Controller Area Network) acquisitions or from datasheets, only simple and poorly accurate models can be obtained. This paper presents an innovative approach that allows avoiding these difficulties: the Genetic Algorithm applied to parameters estimation of powertrain models for the longitudinal dynamic of the vehicle. By employing this technique, parameters estimation is possible even for particularly complex models. The considered approach was tuned and tested in simulation environment showing promising results. Thereafter, the CAN acquired target quantities needed for the minimization, e.g. the engine torque and the wheel speed, and the outputs of the Genetic Algorithm identified model have been compared showing the effectiveness of the proposed approach in real data validation.