Deep Learning-based Human Implantable Nano Molecular Communications

In this paper, we propose a novel nano-molecular communication system, including a nano receiver design and detection strategies. We show how machine intelligence can be incorporated into the practical implementation of nano communications. We introduce a testbed employing a biosensor chip as a receiver. The chip is made to be sufficiently small to be implanted under the human skin with no harm while detecting concentrations of glucose molecules over time. Molecules are released by a transmitter, to convey information through a thin pipe. For this configuration, the channel model is unknown, and the sensor dynamics can differ with according to the manufacturing process. Therefore, it is more desirable to find a universal strategy than using closed-form channel expressions so that it can be less sensitive to the channel and sensor variation. Learning-based approaches are likely to solve the problem. Therefore, in this paper, we suggest detection strategies with and without machine learning. We first describe our intuitions of nanomachine design from observations, and we show how the learning-based techniques can benefit the system by reducing the design burden and enhancing the accuracy of data detection. The study concludes by showing sample results of real data transmission.