Receiver Design in Molecular Communications: An Approach Based on Artificial Neural Networks

Abstract

The design of communication systems typically relies on the development of mathematical models that describe the underlying communication channel. In many communication systems, however, accurate channel models may not be known, or the models may not be accurate enough or even not available for efficient system design. In these scenarios, a completely new approach to communication system design and analysis is required. An example of such situations arises in the emerging research field of molecular communications, for which it is very difficult to develop accurate analytical models for several operating scenarios. In this context, the use of data-driven techniques based on artificial neural networks may provide an alternative and suitable solution towards the design and analysis of molecular communication systems. In this paper, we explore the potential of artificial neural networks for application to the design of robust receiver schemes. We study a molecular communication system in the presence of inter-symbol interference and show that a receiver based on artificial neural networks can be trained by using only empirical (raw) data and can provide the same performance as a receiver that has perfect knowledge of the underlaying channel model.