PPO Algorithm-Assisted Design of Absorptive Common-Mode Suppression Filters

Abstract

In this article, the common-mode suppression filters (CMF) are synthesized using deep reinforcement learning algorithm called proximal policy optimization (PPO). The Latin hypercube is firstly employed to sample the parameters of complementary split-ring resonator structure, which is used as a reflective CMF. Then, with the help of PPO algorithm, a resistor-equipped defected ground structure is developed to absorb the reflective common-mode noise. To prove the feasibility of the proposed optimization method, two samples are designed and tested. The first sample is an absorptive CMF with absorption band located at 5 GHz, and the second sample is with dual absorption bands located at 2.4 GHz and 5 GHz. The results demonstrate that the proposed approach enables the automatic design of CMF.