Abstract:Driving cycles construction, which aims to generate various vehicle driving profiles corresponding to typical traffic conditions, plays an important role in the evaluation of vehicle emissions, economy and mileage. Existing methods usually represent the speed-time distributions of driving data in the space spanned by hand-crafted features, and select typical sequences to combine driving cycle curves. However, since the driving data is treated as static, the inherent dynamic characteristics and temporal dependency tend to be ignored, resulting in low accuracy and insufficient robustness. To address this issue, this paper proposes a dynamic driving cycle construction framework, in which feature extraction and sequence clusters are achieved in an unsupervised joint learning manner. Specifically, the driving data are firstly encoded by a Bi-directional Long Short-Term Memory (BiLSTM) branch to capture the temporal correlation property of driving sequences. Then, a temporal clustering branch is presented to achieve soft distribution clustering of feature sequences by introducing a relative-entropy-based regularization term into the coding unit. The two branches are iteratively updated until stable feature learning and clustering results are obtained. Consequently, each branch benefits from the additional improvement over the previous branch during the iteration process. Finally, typical driving sequences are selected according to the intra-class/extra-class distance and class proportion, and then assembled to generate driving cycles profiles. To verify the performance of our proposed method, evaluations are performed on the on-road driving data of light vehicles in Fuzhou, in which the constructed driving cycle from our methods is substituted into COPERT model to estimate and visualize the road emissions, and the experimental results demonstrate that our proposed methods can greatly improve the accuracy and robustness of the constructed driving cycle.

A Construction Method of Truck Driving Cycles Considering Driving Style

A novel heavy-duty truck driving cycle construction framework based on big data

Developing driving cycles for garbage trucks to estimate fuel consumption

A Normalized Approach for Evaluating Driving Styles Based on Personalized Driver Modeling

A Study of Personalized Shift Strategy Based on Styled Driver Modeling

Low-Cost Reconstruction of Typical Driving Cycles Based on Empirical Information and Low-Frequency Speed Data

Experimental Research on Development Method of Urban Driving Cycle for Passenger Vehicle

Constructing Representative Driving Cycle for Heavy Duty Vehicle Based on Markov Chain Method Considering Road Slope

Developing High-Precision Battery Electric Forklift Driving Cycle with Variable Cargo Weight

A Novel Method of Developing Driving Cycle for Electric Vehicles to Evaluate the Private Driving Habits

Construction and optimization of representative actual driving cycles based on the improved autoencoder

Personalized Driver Based Automatic Shift-Scheduling Optimization for Heavy-duty Truck

Evaluating Driving Styles by Normalizing Driving Behavior Based on Personalized Driver Modeling

Optimization based method to develop representative driving cycle for real-world fuel consumption estimation

Comparing Fuel Consumption Based on Normalised Driving Behaviour: a Case Study on Major Cities in China

A novel optimization-based method to develop representative driving cycle in various driving conditions

Construction of electric vehicle driving cycle for studying electric vehicle energy consumption and equivalent emissions

Multimode Energy Management for Plug-In Hybrid Electric Buses Based on Driving Cycles Prediction

Developing a Regional Drive Cycle Using GPS-Based Trajectory Data from Rideshare Passenger Cars: A Case of Chengdu, China

UJ-FLAC: Unsupervised Joint Feature Learning and Clustering for Dynamic Driving Cycles Construction