Channel Modeling for Mobile Molecular Communication With Anomalous Diffusion by Deep Neural Network

Diffusion-based mobile molecular communication (MMC) systems have shown great potential in nanoscale communication, particularly in the scenarios involving anomalous diffusion. Accurately modeling the anomalous diffusion channel of MMC system with multiple receivers is a challenge. However, prior studies have predominantly addressed conventional analytical approaches to characterize the channel impulse response (CIR) of static molecular communication system under normal diffusion channel. However, the deduction method cannot adapt to time-varying and complex channel conditions. In this paper, we study a three dimensional MMC system with one transmitter and multiple receivers under anomalous diffusion channel. We propose a method based on deep neural network (DNN) to predict the parameters of the CIR of this MMC system. Simulation results demonstrate that the prediction ability of DNN-based model outperforms the recurrent neural networks (RNN) based and the long short-term memory (LSTM) based models in terms of prediction ability under different anomalous diffusion conditions. The DNN-based model can effectively improve the accuracy of predicting the CIR for this MMC system, providing a new approach for channel modeling in MMC systems with anomalous diffusion.