Case Study: Mercedes-Benz speeds virtual sensor development with MathWorks MATLAB and Simulink

Mercedes-Benz Simulates Hardware Sensors with Deep Neural Networks

MathWorks worked with Mercedes-Benz to address the challenge of simulating automotive hardware sensors with deep neural networks for deployment on resource-limited powertrain ECUs. Their previous manual process for converting Python-based QKeras models was slow, error-prone, and could not meet the strict CPU and memory constraints of the automotive microcontrollers.

The solution implemented by MathWorks used Deep Learning Toolbox and Fixed-Point Designer to create an automated workflow for importing, converting, and deploying the neural network models. This resulted in a flexible process that met all performance requirements and accelerated development speed by 600%, enabling Mercedes-Benz to successfully deploy virtual sensors and apply the workflow to other deep learning applications.

Mercedes Benz

Katja Deuschl

AI Developer