Lens Bending with Reinforcement Learning for Reduced Optical Aberrations

Fu, CailingCailingFuStollenwerk, JochenJochenStollenwerkHolly, CarloCarloHolly2022-03-152022-03-152021https://publica.fraunhofer.de/handle/publica/41348010.1364/FIO.2021.FM3C.7To automate optical system design, a Reinforcement Learning algorithm using lens improvement methods as parameter optimization or lens adding is developed. A proof of concept using lens bending is demonstrated for different learning strategies.enalgorithmsgenetic algorithmsoptical aberrationoptical designoptical systems621Lens Bending with Reinforcement Learning for Reduced Optical Aberrationsconference paper