Dataset of 101 Stochastically Generated Driving Cycles for Powertrain Robustness Evaluation
Description
This dataset provides 101 stochastically generated velocity time-series profiles (driving cycles) specifically designed for the robustness evaluation of Fuel Cell Hybrid Electric Vehicles (FCHEVs) and their Energy Management Systems (EMS). Standard certification cycles are deterministic and often fail to capture the high-variance dynamics of real-world driving. To address this, these synthetic profiles were generated using a discrete-time Markov chain stochastic approach trained on established driving databases (NEDC and UDDS). A key feature of this dataset is the integration of a continuous aggressiveness parameter (PSI) ranging from 0.00 (mild, eco-driving) to 1.00 (highly aggressive, worst-case scenarios). The 101 individual profiles systematically sweep this PSI spectrum with a step of 0.01, maximizing acceleration variance to provide a controlled environment for sensitivity analysis. This data is highly valuable for researchers and automotive engineers aiming to test predictive control algorithms, and assess fuel cell or battery degradation under unpredictable, transient stress conditions. Dataset Contents: Data_Files (Folder): 101 MATLAB data files (.mat). Each file contains a 1800-second velocity profile sampled at 1 Hz, ready to be imported as a timeseries object. Summary_Table.xlsx: A comprehensive metadata spreadsheet cross-referencing all cycles and detailing their specific PSI level, maximum velocity, and acceleration variance score. Visualize_Cycle.m: A lightweight MATLAB script provided for quick plotting and visualization of the data. ReadMe.txt: Detailed instructions for loading the data into the MATLAB workspace and integrating it with Simulink.
Files
Steps to reproduce
The dataset was computationally synthesized using a discrete-time Markov chain model. To reproduce similar stochastic profiles, researchers should follow these conceptual steps: Extract velocity-time profiles from standard regulatory cycles (e.g., NEDC, UDDS) to construct empirical Transition Probability Matrices (TPM) based on kinematic state spaces (velocity and acceleration). Perturb the base TPM using a continuous aggressiveness factor (ranging from 0 to 1). Higher factor values should artificially increase the probability of transitioning to higher acceleration states. Utilize Monte Carlo simulations to generate thousands of candidate velocity time-series from the perturbed TPMs. Evaluate the candidate profiles and isolate the one that maximizes the variance of acceleration. Apply a moving average filter to the selected profile and enforce strict kinematic constraints (e.g., maximum acceleration of 3.0 m/s² and deceleration of -4.5 m/s²) to ensure the cycle is physically viable for a heavy-duty Fuel Cell Hybrid Electric Vehicle.