Verification and Deployment of AI-Based Virtual Sensors on Infineon AURIX TC4x

See how to verify and test properties of an AI network used in a real-time system to ensure it can be deployed with confidence. The demo uses an AI virtual sensor to estimate the rotor position of a motor in real-time running on an an Infineon^® AURIX™ TC4x microcontroller. Due to a hysteresis brake connected to the motor, a wide range of operating scenarios can be generated for training, verification and real-time hardware control. MATLAB^® and Simulink^® were used to develop the software on the AURIX, including verifying and deploying the AI virtual sensor.   

Virtual sensors are a powerful AI technique for mimicking the behavior of a physical sensor when the signal of interest cannot be directly measured, or when a physical sensor adds too much cost and complexity to the design. The behavior of this virtual sensor will be verified in several ways including estimating how sensitive the network predictions are to input perturbation, and creating a distribution discriminator that separates data into in- and out-of-distribution. The rigorous verification and testing methods employed in this example are essential for achieving certification under standards such as ISO 26262 and ISO 8800, ensuring the safe and reliable deployment of the AI-based virtual sensor in automotive applications. 

The AI-based software performing the motor control functions demonstrates how this can be done with Model-Based Design while leveraging the unique capabilities of the Infineon AURIX TC4x microcontroller family such as the Parallel Processing Unit (PPU) and Converter Digital Signal Processor (CDSP).