Hybrid method for more reliable virtual sensors in vehicle dynamics control systems


New technical paper entitled “Ensuring the reliability of virtual sensors based on artificial intelligence in vehicle dynamics control systems” from the University of Duisburg-Essen.

“Using virtual sensors in vehicles is a cost-effective alternative to installing physical hardware. In addition to physical models derived from theoretical modelling, artificial intelligence and machine learning approaches are increasingly used, which integrate experimental modelling. Due to the resulting black box characteristics, virtual sensors based on artificial intelligence are not entirely reliable, which can have fatal consequences in safety-critical applications. Therefore, a hybrid method is presented that preserves the reliability of artificial intelligence-based estimates. The example application is the state estimation of the roll angle of the vehicle. The state estimation is coupled to a central predictive vehicle dynamics control. Implementation and validation are performed by co-simulation between IPG CarMaker and MATLAB/Simulink. Using the hybrid method, unreliable estimates by the artificial intelligence-based model resulting from erroneous input signals are detected and processed. Thus, a valid and reliable state estimate is available everywhere.

Find the technical document in free access here. Published in May 2022.

Sieberg, PM; Schramm, D. Ensuring Reliability of Artificial Intelligence-Based Virtual Sensors in Vehicle Dynamics Control Systems. Sensors 2022, 22, 3513. https://doi.org/10.3390/s22093513

Visit the Semiconductor Engineering Technical Article Library here and discover many more academic articles about the chip industry.


About Author

Comments are closed.