Design of a Triple-Band and Ultra-Thin Absorber with Pixel Structure Based on Reinforcement Learning

The pixel structure regards the design area as the binary matrix, which provides a promising view for the high-performance absorber design. This paper presents the design method for metamaterial absorber (MMA) with pixel structure based on proximal policy optimization (PPO), which is a reinforcement learning method. The triple-band and ultra-thin absorber with pixel structure is designed using the proposed method. Comparative study indicates that the PPO converges faster than the genetic algorithm (GA) and particle swarm optimization (PSO). The absorption peak of the designed absorber occurs at 5.200 GHz, 14.425 GHz and 19.225GHz, respectively with ultra-thin profile.