The assessment of the ecological impact of different powertrain concepts is of increasing relevance considering the enormous efforts necessary to limit the global warming effect due to the man-made climate change. Within this contribution, we adopt existing methods for the optimization of electric and hybrid electric powertrains using a vehicle simulation environment and derive a method to identify the ecological potential of different powertrain concepts for a set of technological parameters in the reference year 2030. By optimizing the parametrization for each powertrain concept and by adapting the respective operating behaviour specifically to minimize the ecological impact, a reliable and unbiased comparison is enabled. We use our optimization environment with the Real Ecological Impact as objective function to compare different powertrain concepts on driving profiles that are based on real driving data recorded in Germany. Despite the fact that all of the considered driving profiles contain trips of similar length, their respective optimized powertrain concepts are different. Plug-In Hybrid vehicles achieve the greatest potential for long-range capable vehicles and are least sensitive to different driving profiles.
Driving style classification can be useful for various intelligent vehicle applications and can improve driving comfort in new driveability functions. Within this contribution, we present a rule-based classification algorithm, which uses aggregated speed-depended acceleration data driving profiles from recorded experiments. We observe, that high values in longitudinal and lateral acceleration occur less frequently the greater they are. If a dynamic driving style is present, high values also occur more frequently compared to the average. From that observation we use the occurrence probability of an acceleration data pair as an indicator for driving style. The proposed approach compares the expected average acceleration according to the driving profile with the actual time series data. Based on a sample time series from recorded experiment data, the driving style is then classified into three different classes: calm, moderate and dynamic. Because the approach relies on a limited amount of parameters with low sensitivity, the classification offers high robustness and is not prone to over-fitting. It reaches overall 68.49% accuracy on recorded real life driving data with various driving context.
As an important aspect of today’s efforts to reduce greenhouse gas emissions, the energy demand of passenger cars is a subject of research. Different drivetrain concepts like plug-in hybrid electric vehicles (PHEV) and battery electric vehicles (BEV) are introduced into the market in addition to conventional internal combustion engine vehicles (ICEV) to address this issue. However, the consumption highly depends on individual usage profiles and external operating conditions, especially when considering secondary energy demands like heating, ventilation and air conditioning (HVAC). The approach presented in this work aims to estimate vehicle consumptions based on real world driving profiles and weather data under consideration of secondary demands. For this purpose, a primary and a secondary consumption model are developed that interact with each other to estimate realistic vehicle consumptions for different drivetrain concepts. The models are parametrized by referring to state of the art contributions and the results are made plausible by comparison to literature. The sensitivities of the consumptions are then analysed as a function of trip distance and ambient temperature to assess the influence of the operating conditions on the consumption. The results show that especially in the case of the BEV and PHEV, the trip distance and the ambient temperature are a first-order influencing factor on the total vehicle energy demand. Thus, it is not sufficient to evaluate new vehicle concepts solely on one-dimensional driving cycles to assess their energy demand. Instead, the external conditions must be taken into account for a proper assessment of the vehicle’s real world consumption.
Driving style classification does not only depend on objective measures such as vehicle speed or acceleration, but is also highly subjective as drivers come with their own definition. From our perspective, the successful implementation of driving style classification in real-world applications requires an adaptive approach that is tuned to each driver individually. Within this work, we propose a transfer learning framework for driving style classification in which we use a previously developed rule-based algorithm for the initialization of the neural network weights and train on limited data. Therefore, we applied various state-of-the-art machine learning methods to ensure robust training. First, we performed heuristic-based feature engineering to enhance generalized feature building in the first layer. We then calibrated our network to be able to use its output as a probabilistic metric and to only give predictions above a predefined neural network confidence. To increase the robustness of the transfer learning in early increments, we used a synthetic oversampling technique. We then performed a holistic hyperparameter optimization in the form of a random grid search, which incorporated the entire learning framework from pretraining to incremental adaption. The final algorithm was then evaluated based on the data of predefined synthetic drivers. Our results showed that, by integrating these various methods, high system-level performance and robustness were met with as little as three new training and validation data samples in each increment.
In this paper, we introduce a control method for the linear quadratic tracking (LQT) problem with zero steady-state error. This is achieved by augmenting the original system with an additional state representing the integrated error between the reference and actual outputs. One of the main contributions of this paper is the integration of a linear quadratic integral component into a general LQT framework. In this framework, the reference trajectories are generated using a linear exogenous system. During a simulative implementation for the specific real-world system of a car-in-the-loop (CiL) test bench, we assumed that the ‘real’ system was completely known. Therefore, for model-based control, we could have a perfect model identical to the ‘real’ system. It became clear that for CiL, stable solutions cannot be achieved with a controller designed with a perfect model of the ‘real’ system. On the contrary, we show that a model trained via Bayesian optimization (BO) can facilitate a much larger set of stable controllers. It exhibited an improved control performance for CiL. To the best of the authors’ knowledge, this discovery is the first in the LQT-related literature, which is a further distinctive feature of this work.
Beside others, climate change and digitalization are trends of huge public interest, which highly influence the development process of future powertrain technologies