Abstract

Over the last decade, there has been a growing interest in utilizing novel photonic and optical devices for a diverse range of applications. For the next generation of wireless communication networks, the development of new and optimal optical devices is inevitable. Existing optical network infrastructure cannot meet the stringent requirements of next-generation data networks (such as a 1000-fold increase in bandwidth demand, very low latency, better spectral and energy efficiency, etc.). In other words, the physical layer of the communication network must be revolutionized to provide the proper foundation for these emerging technologies.
Optical networks are based on propagating light. Light propagation in realistic settings is usually a complicated phenomenon. When it comes to the context of optical devices and its propagation in the new devices, the complexity of the problem becomes much higher. In other words, the relations between the light propagation characteristics and the structural parameters of the new devices are mostly unknown. Therefore, the conventional method for designing such devices in the absence of a clear analytic description is usually based on a trial and error process. This method has many disadvantages, being time-consuming, inefficient, and the designed device is usually far from an optimized one. Also, the designing process needs intensive human involvement.
Therefore, to fill this gap, we have utilized artificial intelligence (AI) techniques to design, analyze, and optimize several different optical devices. More specifically, we have proposed several optimization frameworks for designing orbital angular momentum (OAM) fibers, large mode area photonic crystal (PhC) fibers, waveguide-based LP01 to LP0m mode converter, PhC filters, PhC sensors, and PhC-enhanced light-emitting diodes (LEDs). In all of these devices, we are dealing with a complicated system in which the relationships between the structural parameters and the output performance merit factors are very complicated. Such problems have a long simulation runtime, so it is not viable to employ an exhaustive optimization algorithm, which evaluates all of the possible combinations of the parameters to find the optimal one. Therefore, we consider our problem as a black box and use the AI optimization algorithm to find the optimal solution. Eventually, the proposed optimization frameworks open up an effective way to design high-performance optical devices for a diverse range of applications and pave the way for the development of next-generation optical devices for next-generation optical networks.