Machine Learning Algorithms in Mobile Applications for the prediction of unknown states

One of the focuses in automotive research is the development and function enhancement of vehicle dynamics control systems (such as ESP, ABS, ASR, ...). These stabilize the vehicle in critical driving maneuvers by suitable actuation of actuators (for example braking of individual wheels). A precise detection of the current driving condition is indispensable for this. This is ensured by a multitude of different measuring signals (sensor, bus signals). The measurement of the required sizes is sometimes a challenge, both from the feasibility viewpoint as well as from an economical point of view. A science-focused strategy to tackle this problem is the estimation of the required measured variables. To estimate the quantities, already available signals are used, from which the missing quantities are predicted. The physical contexts or models, to what extent missing measurement values ​​can be estimated by known measured variables, are sometimes not known.

Classical control systems are often based on a model-based approach or follow a defined logic. These rules are, however, only limited in their ability to learn or appreciate unknown connections. The estimation of the missing quantities is rather an independent model, which predicts unknown target variables from existing data. This learning and recognition of patterns from existing data is called machine learning and offers the possibility to integrate a "Virtual Sensor" into the real system. This simulated sensor can replace a "Hardware" sensor if the real integration in the vehicle is associated with a high outlay or a high cost.

For the integration of the virtual sensor, the underlying algorithm has to be trained by means of a prototype, so that it can recognize regularities. The training data used for this are obtained from real sensor data. After the training and validation of the model, the physical sensor can be dispensed with and the machine learning algorithm takes over the role of the sensor.

Project supervisor at the Chair of Mechatronics:

• Sebastian Blume
• Sebastian Reicherts
• Philipp Sieberg

Participating Chairs / Departments at the UDE:

• Chair of Mechatronics (Prof. Schramm)

Belongs to the main focus of the Faculty of Engineering at UDE:

• Smart Engineering
• Human-centered Cyber-physical Systems

Link to the faculty (Profile Focal Points): www.uni-due.de/iw/en/research/psp.php