A SIMULATION STUDY ON THE IDENTIFICATION OF ECO-DRIVING BEHAVIOUR.

Eco-driving is considered as one of the effective ways to reduce energy consumption. The objective of this study is to establish an eco-driving behaviour identification model and analyse the eco-driving behaviour characteristics. First, this study built a driving simulation platform and conduct an experiment for research. Then, based on significant driving behaviours related to fuel consumption, Piecewise Linear Representation (PLR) method was used to fit multivariate time series consisting of Significant driving behaviour variables. At last, the features of each time series segment were extracted as the input of a Random Forest (RF) model to recognize the eco-driving behaviour. The results indicated that the depth of the acceleration pedal, the depth of the clutch pedal, the depth of the brake pedal, steering wheel angle, and gear position had significant effects on fuel consumption. The eco-driving behaviour identification model demonstrated a high predictive power with a prediction accuracy of 0.718. The graph of eco-driving behaviours during lane change was established. The conclusions provide theoretical support for developing eco-driving intervenes. (Received in March 2022, accepted in May 2022. This paper was with the authors 2 weeks for 1 revision.) [ABSTRACT FROM AUTHOR]