Deep Learning-Based Estimation of Distance and Emission Time in Mobile Molecular Communication

Abstract

In molecular communication, estimating the distance between the transmitter and the receiver has a variety of applications at the micro-scale, such as the transmitter or the receiver localization, targeted drug delivery, etc. Existing traditional methods for distance estimation typically rely on mathematical derivations that assume prior knowledge of the emission times of molecules from the transmitters. However, when the emission time is unknown, accurately estimating the distance becomes significantly more challenging and involves higher deduction complexity. In this paper, we propose a deep learning-based approach to simultaneously estimate the distances between multiple transmitters and a receiver and the emission times of molecules. We evaluate the prediction performance of the Transformer model. Simulation results indicate that the Transformer model achieves the highest prediction accuracy for both distances and emission times estimation under given error thresholds, outperforming both the Deep Neural Networks and Informer models, which demonstrates strong prediction ability of the Transformer model.