Estimation and detection for molecular MIMO Communications in the internet of bio-nano things

Abstract

For the Internet of Bio-Nano Things (IoBNT) applications demanding high transmission rates, a well-modeled Molecular Communication (MC) channel is essential. The existing studies proposing multiple-input and multiple-output (MIMO) models for MC, however, often make the unrealistic assumption of using ideal receivers with perfect absorption. Hence, this paper proposes a molecular MIMO channel model with spherical transmitters and partially-absorbing ligand receptor-based receivers underpinned by four unique parameters. For the non-analytical nature of the MIMO channel, we use a supervised learning algorithm to estimate the number of molecules in the reception space. We evaluate the root mean square error (RMSE) of our solution, which returns consistent results. The estimation is used for ligand-receptor binding statistics, in which the intersymbol inference (ISI) and molecular interference are considered. We also propose two techniques based on convolutional and recurrent neural networks (CNN & RNN) as alternatives to the generic threshold-based detection. Our detectors outperform the threshold-based technique; specifically, the CNN-based method improves the mean bit error rate (BER) performance three times.