Cloud-based Digital Twin Solution for Predictive e-Mobility Applications

Autoren

L. Schäfers, S. Pischinger - TME, RWTH Aachen University, M. Wegener, R. Savelsberg, A. Alberti, G. Cecchini - FEV, K. Badalian - MMP, RWTH Aachen University

Zusammenfassung

Predictive driving functions utilizing long-horizon load profile prediction and digital twin simulations can make an important contribution to further reduce energy consumption of road vehicles and improve the overall user experience. In this context, the following paper presents an infrastructure and functions framework that efficiently enables the well-investigated potentials and advantages of predictive driving. A set of hardware including a real-time capable on-board unit, a cloud computing instance, a vehicle communication data logger and a HMI device permits fully automated extension of the vehicle’s storage and computing capacity as well as data availability. In the cloud, vehicle data is combined with anonymized driver data and third-party traffic, route and weather information in order to train data-driven vehicle- and driver-related prediction algorithms such as driving destination forecasts, vehicle resistance estimation or speed profile prediction. These models are then used within a simulation framework to onlineforecast the load profile, enabling the implementation of predictive driving functions such as predictive thermal management, residual range prediction or charge strategy planning. The automation of the process chain and the training of the models over night off-board in the cloud environment allows a continuous improvement of the prediction accuracy with every new driven trip. Additionally, the scalability in the cloud environment enables the execution of computationally intensive algorithms like a simulationbased range polygon that requires multiple executions of a range prediction algorithm towards various driving directions.