Rapid Prototyping of Machine Learning Systems

Machine learning systems are gaining acceptance in the fields of inferential sensing, mechatronic control and prognostics. However, software implementations can place excessive demands on the ECU, and so real-time classification rates are not always possible.

This paper describes the integration of a hardware implementation of a machine learning algorithm into a comprehensive hardware and software prototyping environment for powertrain applications. The paper describes the hardware and software architectures developed, provides an overview of the new methodologies necessary to access the power of the machine learning system, and illustrates its application in the powertrain control field