Browse Publications Technical Papers 2014-01-2091

Vibration Comfort Control for HEV Based on Machine Learning 2014-01-2091

Hybrid electric vehicles (HEVs) with a power-split system offer a variety of possibilities in reduction of CO2 emissions and fuel consumption. Power-split systems use a planetary gear sets to create a strong mechanical coupling between the internal combustion engine, the generator and the electric motor. This concept offers rather low oscillations and therefore passive damping components are not needed. Nevertheless, during acceleration or because of external disturbances, oscillations which are mostly influenced by the ICE, can still occur which leads to a drivability and performance downgrade.
This paper proposes a design of an active damping control system which uses the electric motor to suppress those oscillations instead of handling them within the ICE control unit. The control algorithm is implemented as part of an existing hybrid controller without any additional hardware introduced. Because the system is rather slow and acting upon detection of oscillations has no reasonable effect, the controller has to predict the future behavior and the torque distribution in the drive-train. Based on this prediction, the controller generates a damping torque signal for the electric motor which provides an additional torque to suppress upcoming oscillations. The proposed controller uses a machine learning method, Q-Learning to train the controller fully automatically. The target of the control is to increase drivability without affecting the performance of the vehicle. Implementation of the proposed controller can dramatically reduce the complexity of the engine control unit which leads to a lower development price.


Subscribers can view annotate, and download all of SAE's content. Learn More »


Members save up to 16% off list price.
Login to see discount.
Special Offer: Download multiple Technical Papers each year? TechSelect is a cost-effective subscription option to select and download 12-100 full-text Technical Papers per year. Find more information here.