Abstract
Research and/or Engineering Questions/ Objective To enhance fuel-efficiency through reduction of transient energy loss resulting from unnecessary EV/HEV mode changes, driving patterns of HEVs are classified and the control methods that come with it are developed. Methodology The key factors reflecting driving characteristics of vehicles are selected and the driving patterns are decided by comparative analyses of various machine learning techniques. The consequential driving patterns are classified per each feature such as road types and driving styles. Also, multiple control methods corresponding to the classified patterns are proposed for improving energy economy. Further, with the development of rules for soft driving pattern transition, performance of the proposed strategy is analyzed on real-world driving tests. Results The simulation and experiment tests indicate that the driving pattern classification/decision developed by SVM (Support Vector Machine) training on the basis of off-line data illustrates the improved performance for energy efficiency. Limitations of this study Since the research assumes that the additional informative inputs such as look-ahead road/route information and driving style are not given, there exist cases where the decision timing is hard to unceasingly track abrupt changes for road circumstances and driving style. What study does the paper offer that is new in the field including in comparison to other work by the authors? The paper provides a reasonable possibility that the machine learning-based driving pattern decision contributes to the enhancement of energy efficiency for HEV on real-world driving by only using basic driving factors. Conclusions By utilizing machine learning methods, a driving pattern decision and control strategy to improve energy efficiency of HEV has been developed. Various on-line/off-line tests show that the strategy retains energy-efficient effects.