Deploy AI-Based Functions into Rapid Prototyping for Real-Time Applications

New functions based on machine learning models are being developed across the automotive industry for multiple applications with the goal of improving vehicle performance and customer satisfaction. To study the viability and perform the calibration of such functions, they can be deployed into rapid prototyping units to execute in-vehicle validations. However, the deployment of models coming from newer development frameworks into real-time applications can bring some challenges using the classical toolchain approach.

This presentation shows how Hyundai Motor Europe (in cooperation with MathWorks) uses Simulink^® code generation and the target library for third parties to deploy different machine learning models into a defined toolchain for rapid control prototyping and ECU calibration. Together, we have established a workflow including different solutions to run models designed and trained within MATLAB^® or PyTorch^® frameworks. This gives the team great flexibility to extend its research and development capabilities across different environments.