Towards a deep reinforcement learning integration into model-based systems engineering

The integration of Deep Reinforcement Learning (DRL) in Model-Based Systems Engineering (MBSE) is a promising approach that can lead to significant benefits for system designers and developers. DRL is a branch of machine learning where an agent learns to make decisions by interacting with an environment, receiving feedback in the form of rewards or punishments that indicate the quality of its actions, and adjusting its decision-making policy to maximize the cumulative reward over time. MBSE provides a structured approach to system design, which can help to clarify system requirements, identify potential issues, and improve the overall efficiency of the system development process. This model-based approach can be particularly useful for DRL, which requires a clear understanding of the system environment and objectives to develop the system’s behavior. We propose a method for integrating DRL into MBSE, where the desired system behavior is defined in a model-based representation using a modeling language to describe the relevant design components for DRL. The method's model framework is applied and evaluated to an example use case using SysML as the modeling language. This integration enables system designers to use DRL with the benefits and support of MBSE.