Framework for Automated Development and Modelling of Complex Photonic Systems

The complexity of photonic systems is increasing as a result of system miniaturization and the functional integration of components. Additionally, novel materials, processes, tools, and methods pose further challenges for developers. These challenges are often addressed through intuition or experience, which can lead to design approaches that prematurely exclude parts of the solution space without prior consideration. To avoid this, we propose an automated framework for the development and modeling of constraint-based optical systems. To meet the growing demand for cross-domain functionality of components, a holistic development approach is recommended. This approach is enabled by deriving constraints directly from the requirements. These constraints restrict the solution space to feasible designs. Within this reduced solution space, each possible design is optimized and evaluated against the others using a loss function. This methodology enables the unbiased development of optical systems. Furthermore, the framework can be embedded in a multi-agent system. Various agents take on tasks such as simulating the different designs or solving the constraint problems. The advantages here are that the agents negotiate and resolve conflicting goals (e.g., contradictory constraints) themselves. This defines an AI-based framework for constraintbased and unbiased development of optical systems.