Marius Wegener
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Marius Wegener has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Jakob Andert
Chair of Mechatronics in Mobile Propulsion, Faculty of Mechanical Engineering, RWTH Aachen University, 52062 Aachen, Germany
8 shared publications - Lucas Koch
Teaching and Research Area Mechatronics in Mobile Propulsion, RWTH Aachen University, Aachen, Germany
4 shared publications - Kevin Badalian
Teaching and Research Area Mechatronics in Mobile Propulsion, RWTH Aachen University, Aachen, Germany
3 shared publications - Florian Herrmann
FEV Europe GmbH Aachen Nordrhein‐Westfalen Germany
2 shared publications - Mario Picerno
Teaching and Research Area Mechatronics in Mobile Propulsion, RWTH Aachen University, Aachen, Germany
2 shared publications Markus Eisenbarth
Mechatronic Systems for Combustion Engines, RWTH Aachen University, Aachen, Germany
2 shared publications
Latest Publications
- 28 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 74 (7) , 10332-10343
Transfer of Reinforcement Learning-Based Powertrain Controllers From Model- to Hardware-in-the-Loop
Developing powertrain control functions is time-consuming and resource-intensive, often leading to sub-optimal solutions. Reinforcement Learning (RL) allows agents to perform complex control tasks with minimal human involvement, but is often confined to simulations due to testing costs and safety concerns. To effectively apply RL in embedded powertrain control, agents must be able to handle real-world scenarios, particularly through direct interaction with real actuators and control systems. Therefore, this research applies Transfer Learning (TL) and X-in-the-Loop (XiL) simulations to develop agents that can seamlessly transition and perform robustly in real-world environments. For transient exhaust gas re-circulation control of an internal combustion engine, the process begins with a computationally inexpensive Model-in-the-Loop (MiL) simulation to select a suitable algorithm, fine-tune hyperparameters, and conduct preliminary training. In the next step, pre-trained agents are transferred to an advanced Hardware-in-the-Loop (HiL) system with real hardware using TL for further training. Compared to agents trained entirely on HiL systems, transferred agents required significantly less real-world training time (up to 5.9 times shorter) while outperforming the series production Engine Control Unit (ECU). The results highlight that for real-world effectiveness, integrating actual hardware into training is essential, reward fine-tuning plays a critical role in optimizing these interactions, and the maturity of the policy significantly influences both training duration and overall performance.
- 28 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Vehicular Technology
- Vol. 74 (7) , 10332-10343
Transfer of Reinforcement Learning-Based Powertrain Controllers From Model- to Hardware-in-the-Loop
Developing powertrain control functions is time-consuming and resource-intensive, often leading to sub-optimal solutions. Reinforcement Learning (RL) allows agents to perform complex control tasks with minimal human involvement, but is often confined to simulations due to testing costs and safety concerns. To effectively apply RL in embedded powertrain control, agents must be able to handle real-world scenarios, particularly through direct interaction with real actuators and control systems. Therefore, this research applies Transfer Learning (TL) and X-in-the-Loop (XiL) simulations to develop agents that can seamlessly transition and perform robustly in real-world environments. For transient exhaust gas re-circulation control of an internal combustion engine, the process begins with a computationally inexpensive Model-in-the-Loop (MiL) simulation to select a suitable algorithm, fine-tune hyperparameters, and conduct preliminary training. In the next step, pre-trained agents are transferred to an advanced Hardware-in-the-Loop (HiL) system with real hardware using TL for further training. Compared to agents trained entirely on HiL systems, transferred agents required significantly less real-world training time (up to 5.9 times shorter) while outperforming the series production Engine Control Unit (ECU). The results highlight that for real-world effectiveness, integrating actual hardware into training is essential, reward fine-tuning plays a critical role in optimizing these interactions, and the maturity of the policy significantly influences both training duration and overall performance.
- Open Access
- 26 June 2024
- Published by Springer Nature in Applied Intelligence
- Vol. 54 (17-18) , 8384-8398
LExCI: A framework for reinforcement learning with embedded systems
Advances in artificial intelligence (AI) have led to its application in many areas of everyday life. In the context of control engineering, reinforcement learning (RL) represents a particularly promising approach as it is centred around the idea of allowing an agent to freely interact with its environment to find an optimal strategy. One of the challenges professionals face when training and deploying RL agents is that the latter often have to run on dedicated embedded devices. This could be to integrate them into an existing toolchain or to satisfy certain performance criteria like real-time constraints. Conventional RL libraries, however, cannot be easily utilised in conjunction with that kind of hardware. In this paper, we present a framework named LExCI, the Learning and Experiencing Cycle Interface, which bridges this gap and provides end-users with a free and open-source tool for training agents on embedded systems using the open-source library RLlib. Its operability is demonstrated with two state-of-the-art RL-algorithms and a rapid control prototyping system.
- 26 June 2024
- Published by Springer Nature in Applied Intelligence
- Vol. 54 (17-18) , 8384-8398
Open AccessLExCI: A framework for reinforcement learning with embedded systems
Advances in artificial intelligence (AI) have led to its application in many areas of everyday life. In the context of control engineering, reinforcement learning (RL) represents a particularly promising approach as it is centred around the idea of allowing an agent to freely interact with its environment to find an optimal strategy. One of the challenges professionals face when training and deploying RL agents is that the latter often have to run on dedicated embedded devices. This could be to integrate them into an existing toolchain or to satisfy certain performance criteria like real-time constraints. Conventional RL libraries, however, cannot be easily utilised in conjunction with that kind of hardware. In this paper, we present a framework named LExCI, the Learning and Experiencing Cycle Interface, which bridges this gap and provides end-users with a free and open-source tool for training agents on embedded systems using the open-source library RLlib. Its operability is demonstrated with two state-of-the-art RL-algorithms and a rapid control prototyping system.
- 30 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 24 (12) , 15066-15076
Adaptive Traffic Light Control With Deep Reinforcement Learning: An Evaluation of Traffic Flow and Energy Consumption
Cities of all sizes around the world are facing increasing levels of congestion, which leads to increasing travel time and emissions, and ultimately affects the quality of life. Relevant research suggests that adaptive traffic light control systems can improve the traffic flow, but their impact on energy-efficiency of vehicle propulsion systems is not well understood. In this study, we use Proximal Policy Optimization, a Deep Reinforcement Learning algorithm, to develop an optimized adaptive traffic light control systems that controls three traffic lights simultaneously. For this purpose, we have created a microscopic traffic simulation of the city of Aachen, Germany, calibrated on the basis of traffic measurements, where the actual traffic light schedule, a green-wave, fixed-time control scheme, serves as a reference. The traffic simulation is coupled with detailed, physics-based powertrain models of both conventional and electric vehicles, which are validated against chassis dynamometer measurements. By analyzing the complex interactions between traffic light control, the resulting vehicle trajectories and the powertrain components, we show that Reinforcement Learning-based adaptive control can significantly improve the traffic flow, with a 41% increase in average velocity, without any drawbacks in CO2 emission (-1%). Furthermore, we find that maximizing traffic flow and minimizing CO2 emissions are not necessarily contradictory objectives, and identify an increased energy saving potential at low traffic densities. Thus, we prove that adaptive traffic light control can make traffic not only more time-efficient, but also more sustainable.
- 30 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 24 (12) , 15066-15076
Adaptive Traffic Light Control With Deep Reinforcement Learning: An Evaluation of Traffic Flow and Energy Consumption
Cities of all sizes around the world are facing increasing levels of congestion, which leads to increasing travel time and emissions, and ultimately affects the quality of life. Relevant research suggests that adaptive traffic light control systems can improve the traffic flow, but their impact on energy-efficiency of vehicle propulsion systems is not well understood. In this study, we use Proximal Policy Optimization, a Deep Reinforcement Learning algorithm, to develop an optimized adaptive traffic light control systems that controls three traffic lights simultaneously. For this purpose, we have created a microscopic traffic simulation of the city of Aachen, Germany, calibrated on the basis of traffic measurements, where the actual traffic light schedule, a green-wave, fixed-time control scheme, serves as a reference. The traffic simulation is coupled with detailed, physics-based powertrain models of both conventional and electric vehicles, which are validated against chassis dynamometer measurements. By analyzing the complex interactions between traffic light control, the resulting vehicle trajectories and the powertrain components, we show that Reinforcement Learning-based adaptive control can significantly improve the traffic flow, with a 41% increase in average velocity, without any drawbacks in CO2 emission (-1%). Furthermore, we find that maximizing traffic flow and minimizing CO2 emissions are not necessarily contradictory objectives, and identify an increased energy saving potential at low traffic densities. Thus, we prove that adaptive traffic light control can make traffic not only more time-efficient, but also more sustainable.
- Open Access
- 23 May 2023
- Published by Institution of Engineering and Technology (IET) in IET Intelligent Transport Systems
- Vol. 17 (11) , 2154-2170
Centralized model‐predictive cooperative and adaptive cruise control of automated vehicle platoons in urban traffic environments
Recent research has demonstrated that cooperative and automated driving functions can improve traffic efficiency and safety. Traffic efficiency can be increased and the overall vehicle energy consumption can be reduced through cooperative driving functions not only on the highway, but also in urban areas. To this end, this paper presents a linear model predictive control (MPC) algorithm for the optimization of velocity trajectories of vehicles in a platoon in an urban traffic environment. A novel centralized approach is implemented to optimize the total energy consumption of all vehicles while improving traffic efficiency. To solve the optimal control problem, for the first time, an additional data‐driven velocity prediction model of the vehicle in front of the platoon is used to consider the influence of realistic prediction errors when evaluating the developed control strategy in a vehicle simulation environment. The simulated total energy consumption is compared with a rule‐based cooperative adaptive cruise control (CACC) algorithm. Furthermore, the traffic density and the platoon string stability, the robustness and computer time of the centralized approach are also included in the evaluation. Simulation results indicate that, considering all evaluation criteria, the centralized cooperative control strategy can improve energy efficiency by up to 15% in the investigated urban scenario.
- 23 May 2023
- Published by Institution of Engineering and Technology (IET) in IET Intelligent Transport Systems
- Vol. 17 (11) , 2154-2170
Open AccessCentralized model‐predictive cooperative and adaptive cruise control of automated vehicle platoons in urban traffic environments
Recent research has demonstrated that cooperative and automated driving functions can improve traffic efficiency and safety. Traffic efficiency can be increased and the overall vehicle energy consumption can be reduced through cooperative driving functions not only on the highway, but also in urban areas. To this end, this paper presents a linear model predictive control (MPC) algorithm for the optimization of velocity trajectories of vehicles in a platoon in an urban traffic environment. A novel centralized approach is implemented to optimize the total energy consumption of all vehicles while improving traffic efficiency. To solve the optimal control problem, for the first time, an additional data‐driven velocity prediction model of the vehicle in front of the platoon is used to consider the influence of realistic prediction errors when evaluating the developed control strategy in a vehicle simulation environment. The simulated total energy consumption is compared with a rule‐based cooperative adaptive cruise control (CACC) algorithm. Furthermore, the traffic density and the platoon string stability, the robustness and computer time of the centralized approach are also included in the evaluation. Simulation results indicate that, considering all evaluation criteria, the centralized cooperative control strategy can improve energy efficiency by up to 15% in the investigated urban scenario.
- 21 September 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Vehicles
- Vol. 8 (1) , 204-215
Longitudinal Vehicle Motion Prediction in Urban Settings With Traffic Light Interaction
Predictive cruise control functions designed to reduce the energy consumption of intelligent and automated vehicles require an accurate prediction of the upcoming traffic situation in general and the preceding vehicle in particular. This article presents the implementation of prediction models which do not rely on explicit information of the entire preceding vehicle queue. Using this assumption, prediction algorithms based on Conditional Linear Gauss (CLG) models and Deep Neural Network (DNN) were trained using real-world measurements in an urban setting. The training was conducted with 6.6 hours of driving data specifically collected for this study. The results show that both approaches can provide accurate estimates of a preceding vehicles motion. For the CLG models, a queue-estimation logic was implemented to improve the prediction accuracy. The most accurate prediction could be obtained when DNNs are set up with Long Short-Term Memory (LSTM) layers, capable of modeling time dependencies. The results also show that accurate short and mid term predictions of a vehicles trajectory can be made based primarily on sensor data, even when no or only little data from vehicle-to-vehicle (V2V) communication is available.
- 21 September 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Vehicles
- Vol. 8 (1) , 204-215
Longitudinal Vehicle Motion Prediction in Urban Settings With Traffic Light Interaction
Predictive cruise control functions designed to reduce the energy consumption of intelligent and automated vehicles require an accurate prediction of the upcoming traffic situation in general and the preceding vehicle in particular. This article presents the implementation of prediction models which do not rely on explicit information of the entire preceding vehicle queue. Using this assumption, prediction algorithms based on Conditional Linear Gauss (CLG) models and Deep Neural Network (DNN) were trained using real-world measurements in an urban setting. The training was conducted with 6.6 hours of driving data specifically collected for this study. The results show that both approaches can provide accurate estimates of a preceding vehicles motion. For the CLG models, a queue-estimation logic was implemented to improve the prediction accuracy. The most accurate prediction could be obtained when DNNs are set up with Long Short-Term Memory (LSTM) layers, capable of modeling time dependencies. The results also show that accurate short and mid term predictions of a vehicles trajectory can be made based primarily on sensor data, even when no or only little data from vehicle-to-vehicle (V2V) communication is available.
- 8 August 2019
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 234 (4) , 992-1005
Energy saving potentials of modern powertrains utilizing predictive driving algorithms in different traffic scenarios
In this article, we analyze the interaction between powertrain technology, predictive driving functionalities, and inner-city traffic conditions. A model predictive velocity control algorithm is developed that utilizes dynamic traffic data as well as static route information to optimize the future trajectory of the considered ego-vehicle. This controller is then integrated into a state-of-the-art simulation environment for automated driving functionalities to calculate energy saving potentials for vehicles with a conventional gasoline engine powertrain and a P3-hybrid powertrain configuration as well as for a battery electric vehicle based on real driving measurements. The comparison of these powertrains under various traffic conditions shows that all three technologies profit from predictive driving functionalities. The determined reduction in energy demand ranges from 15% to more than 40%, but it is highly dependent on the boundary conditions and the selected powertrain technology. Specifically, it is shown that electrified powertrains can profit the most when the time-gap to the preceding vehicle is maintained at a high level. For a conventional powertrain, this effect is less pronounced and can be attributed to the efficiency characteristics of gasoline engines. It can be concluded that the development of advanced predictive driving functionalities requires microscopic simulation of inner-city traffic to achieve optimum results with regard to energy consumption.
- 8 August 2019
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 234 (4) , 992-1005
Energy saving potentials of modern powertrains utilizing predictive driving algorithms in different traffic scenarios
In this article, we analyze the interaction between powertrain technology, predictive driving functionalities, and inner-city traffic conditions. A model predictive velocity control algorithm is developed that utilizes dynamic traffic data as well as static route information to optimize the future trajectory of the considered ego-vehicle. This controller is then integrated into a state-of-the-art simulation environment for automated driving functionalities to calculate energy saving potentials for vehicles with a conventional gasoline engine powertrain and a P3-hybrid powertrain configuration as well as for a battery electric vehicle based on real driving measurements. The comparison of these powertrains under various traffic conditions shows that all three technologies profit from predictive driving functionalities. The determined reduction in energy demand ranges from 15% to more than 40%, but it is highly dependent on the boundary conditions and the selected powertrain technology. Specifically, it is shown that electrified powertrains can profit the most when the time-gap to the preceding vehicle is maintained at a high level. For a conventional powertrain, this effect is less pronounced and can be attributed to the efficiency characteristics of gasoline engines. It can be concluded that the development of advanced predictive driving functionalities requires microscopic simulation of inner-city traffic to achieve optimum results with regard to energy consumption.
- 28 May 2018
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 233 (5) , 1320-1330
A simulation-based case study for powertrain efficiency improvement by automated driving functions
An increasing level of driving automation and a successive electrification of modern powertrains enable a higher degree of freedom to improve vehicle fuel efficiency and reduce pollutant emissions. Currently, both domains themselves, driving automation as well as powertrain electrification, face the challenge of a rising development complexity with extensive use of virtual testing environments. However, state-of-the-art virtual testing environments typically strictly focus on just one domain and neglect the other. This paper shows the results of a simulation-based case study considering both domains simultaneously. The influence of energy saving automated functionalities on a conventional, a hybrid, and a pure electric powertrain is investigated for a carefully selected inner-city driving scenario. The vehicle simulation models for the different powertrain configurations are calibrated using test bench results and vehicle measurements. A model predictive acceleration controller is developed for realizing the speed optimization function. By considering traffic conditions such as traffic light schedules and a preceding vehicle as the boundary conditions, unnecessary accelerations and decelerations are avoided to reduce the energy demand. The case study is realized by applying this function to the three powertrains variants. As a final result, a clear difference in energy demand is observed: the hybrid powertrain benefits the most in terms of energy demand reduction in the given use case. The results clearly underscore that in future vehicle development programs, the powertrain and the real-world driving functionalities have to be optimized simultaneously to minimize the energy demand during everyday vehicle operation.
- 28 May 2018
- Published by SAGE Publications in Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering
- Vol. 233 (5) , 1320-1330
A simulation-based case study for powertrain efficiency improvement by automated driving functions
An increasing level of driving automation and a successive electrification of modern powertrains enable a higher degree of freedom to improve vehicle fuel efficiency and reduce pollutant emissions. Currently, both domains themselves, driving automation as well as powertrain electrification, face the challenge of a rising development complexity with extensive use of virtual testing environments. However, state-of-the-art virtual testing environments typically strictly focus on just one domain and neglect the other. This paper shows the results of a simulation-based case study considering both domains simultaneously. The influence of energy saving automated functionalities on a conventional, a hybrid, and a pure electric powertrain is investigated for a carefully selected inner-city driving scenario. The vehicle simulation models for the different powertrain configurations are calibrated using test bench results and vehicle measurements. A model predictive acceleration controller is developed for realizing the speed optimization function. By considering traffic conditions such as traffic light schedules and a preceding vehicle as the boundary conditions, unnecessary accelerations and decelerations are avoided to reduce the energy demand. The case study is realized by applying this function to the three powertrains variants. As a final result, a clear difference in energy demand is observed: the hybrid powertrain benefits the most in terms of energy demand reduction in the given use case. The results clearly underscore that in future vehicle development programs, the powertrain and the real-world driving functionalities have to be optimized simultaneously to minimize the energy demand during everyday vehicle operation.