Target Detection and Localization for Mobile Molecular Communication by Deep Learning Methods

Mobile molecular communication (MMC) has various promising applications, such as environmental monitoring, targeted drug delivery and artificial biointelligence. However, the dynamic nature of MMC poses significant challenges for accurate target detection and localization. For the practical applications in MMC, simultaneous target detection and localization with higher accuracy are required. In this paper, we propose a deep learning-based approach to address the issue of target detection and localization in MMC system with multiple target nodes and multiple nanomachines (NMs). We employ a neural network enhanced with attention mechanisms which is specifically a Transformer-based model to learn the complex patterns in molecular diffusion and reception. The data is generated according to the received sequences which are composed of the numbers of molecules arrived at NMs in each time slot by accurately simulating the movement of the targets, NMs and molecules. We have conducted simulation experiments for target detection and localization under both static and mobile conditions. Simulation results indicate that our proposed method performs best in terms of the predicted target detection and localization accuracy compared with other deep learning models including deep neural networks (DNN) and Informer-based especially in mobile scenarios.