Virtual Vehicle

Modeling and Advanced Control

Model-based Design

High product requirements and legislative constraints – e.g. vehicle emission standard – demand a permanent improvement of numerous vehicle components. Due to ascending complexity, further development can be often achieved. Beside the quantity of components, the amount of possible operational statues and various transitions increase. The resulting dependences and interactions of components are multiplied, which have a negative effect on the duration of the development period. Simultaneously, shorter development cycles in the automobile industry demand further improvement of efficiency in both function development and calibration.

The closed-loop and open-loop control system of components essentially determines the performance and takes central importance in modern vehicles. To control the growing complexity, the function development tends to the deployment of plant models based on development methods.

These so-called model-based methods of development advance and/or simplify the function development and parameterization by usage of virtual test rigs and test runs. The objective is the development and implementation of a model-based methodology for creating control systems for processes in automotive applications. The creation of a control system contains the structure's architecture, the parameter setting and the required tools.


Exemplary application of model-based methodology for control systems design:

  • Automatic control of the air pathway of a diesel engine for commercial vehicles. After simulation and virtual implementation the results of the test rig has been also validated under real conditions.
  • Automatic control of shifting and starting operations in an electrically controlled gear of a hybrid car.

Advanced Control

Trends in the development
One possible approach for a successful solution of a control task is the application of a classical PID control concept. The simply intuitive settings and the low computing effort have led to a wide dissemination of that approach.

The progressive enhancement in the complexity of modern vehicles and the increased demands in the past led to a huge rise of necessary adaptations of the original lean control concepts. The result is an over-proportional increase in expenses regarding the function development, testing and parameter setting of the control algorithms.

In order to guarantee an efficient development process with acceptable development effort and corresponding product manufacturing times in the future - beside optimization of classical control concepts - more new control concepts in use for control units are examined.


A key for applying this new procedure is to shift the control parameter setting from the test rig to the virtual test rig.


Model Approximation

A virtual development is state of the art, with development and simulation of models for components. Due to the often high degree of detail and the associated high computational effort, these models are not suitable for the model-based control systems design, as well as for real-time applications or optimization approaches. Simpler descriptions of characteristic relations are necessary to meet the requirements of this model-based method.


In addition to model-order reduction (MOR), where the model describing equation system is reduced, data-driven methods of modeling provide an alternative. Based on measured data an identification of essential input-output relations takes place here. The research focus is on universally applicable methods for efficient creation of fast executable models. Along with the model simplification are sensitivity analysis, design of experiments (DoE) and final evaluation of the model quality.


Modern Control System Concepts

The cross-divisional lead theme control engineering is used as storage and deployment of control engineering know-how and specialized tools as well as the collection of relevant literature.


In addition to already well-established classical concepts such as PID control systems, model-based approaches, such as the flatness-based control, Model Predictive (MPC) and sliding mode control (SMC) are investigated. Virtual sensors, available via model-based observer, increase the applicability of the control rules.


Current Application

Currently, solutions for different control engineering problems in funded co-operation projects are worked out.

At the moment the creation of a modular and easily configurable multi-variable control of the air pathway of a diesel engine is taking place.

In addition, in another project based on a drive train model the control system of a shifting and starting operations of a hybrid vehicle is presented modular. Taking into account the modeled expected dynamic behavior of the combustion engine and with the aid of the electric motor, the control system is optimized.

Current Applications

Increased Safety by motion-prediction pedestrian and robust motion planning:

We work on (predictive) motion planning algorithms for uncertain and dynamic environments. For safety reasons, it is desirable to predict the movements of all traffic participants (especially pedestrians and vehicles). To find safe trajectories in dynamic environments, we use advanced Machine Learning Approaches and new motion planning algorithms.


Generic real-time modeling and model-based control of complex transmission topologies:

Model-based approaches are of huge interest for automotive industry to handle increasing number and complexity of electrified drivetrains in combination with increasing requirements to the shifting performance. To come up to this trend a current research project focuses on modeling and model-based control of complex transmission topologies. Hereby real-time capability plays a key role in order to support subsequent application on transmission control unit. Real-time capability requires fixed time step solution of the model equations. Solving model equations containing friction elements (e.g. clutches) is a major challenge, since locking friction elements cause a system to vary its order and structure. A generic modeling approach for transmission topologies containing multiple friction elements is developed, supporting all-purpose usability. Based on this modeling approach an embedded observer is designed. Such observers facilitate the control of gear shifts by providing virtual measurement data. The generic modeling approach is further used to design a model-based control strategy for the control of gear shifting, focusing on complex drivetrain topologies with multiple propulsion elements (e.g. hybrid electric vehicles).The challenge is to enable optimal performance in the area of conflict between shifting comfort and energy efficiency.


Energy efficient driving:

Many vehicles, some of which are already in a serial production, are equipped with actuators and sensors necessary for precise control of both longitudinal and lateral vehicle movement. Beside sensors for perception of the environment, vehicles are also equipped with sensors for measuring internal states of a vehicle such as battery states, power consumptions, etc. This allows controlling vehicle based on much more information than a human driver is using.

This work studies the problem of autonomous driving from energy consumption point of view.  Using available trip information, onboard sensor information and known vehicle model the controller should control vehicle movement using the least energy for desired transportation task. Controller should be based on computationally efficient optimization methods customized for vehicle driving tasks required to achieve real-time operation.