Parameter Optimization of Impedance Gradient Change Medium Based on Reinforcement Learning

摘要

Based on the relative researches, in order to solve the problem that the parameters of impedance gradient change medium are difficult to be optimized and generalized in different environments, an optimization method of the parameters of the impedance gradient change medium based on reinforcement learning was proposed. First, the propagation principle of sound wave in impedance gradient medium was analyzed. The sound field distribution in the medium was also studied, in order to master its acoustic characteristics. Second, the parameters of sound velocity and impedance distribution were optimized by DQN algorithm to reduce the sound reflection. Finally, the effectiveness of the proposed reinforcement learning model was verified by the traditional method. The experimental results show that the method presented in this paper was superior to the traditional method. The trained parameters are effective to reduce the acoustic reflection to a lower level.