Improving the efficiency of system development is essential for quickly addressing rapid changes in social conditions, which is why data-driven system design methods based on machine learning are expected to serve as a means to achieve this. In order to further enhance control stability without relying on the nature of data, we proposed a method to learn the control object with a specially structured machine learning model. When using this model as a predictor, solutions to mathematical problems that require optimal control under various conditions (optimal control problems) are mathematically guaranteed to be unique and continuous. Currently, we are working to apply the method to engine control in order to achieve smooth control while improving acceleration and environmental performance.~{*1}
*1: Jointly developed in part with Toyota Industries Corporation
Machine learning model for developing highstability control systems