Resource Allocation Optimization in Mobile Multiuser Molecular Communication by Deep Neural Network

Mobile molecular communication (MMC) is expected to be a promising technology for drug delivery. This paper studies a multiuser MMC system in a three-dimensional diffusive environment, which is composed of multiple transmitter nanomachines and one receiver nanomachine. Considering that all transmitter nanomachines release the same type of molecules for information transmission, the mechanism of time division multiple access (TDMA) is employed in this system. Under the release resource constraint which requires that the total number of released molecules of all transmitter nanomachines is fixed, the resource allocation optimization plays a significant role in the performance of this system. When the environmental variables in this multiuser MMC system change, the traditional optimization algorithms need to reoptimize the resource allocation to minimize the average bit error probability (BEP) of this system, which results in more run time. In order to reduce the run time, we propose an algorithm designed based on deep neural network (DNN) to obtain the optimal resource allocation scheme. For the trained DNN, once the input is given, it does not need to re-execute the optimization process and the output can be instantaneously obtained. The numerical results show that the proposed algorithm has a shorter run time and lower average BEP compared with other existing traditional optimization algorithms used in MMC, including bisection algorithm and genetic algorithm. The optimization results are approximate to the optimal solutions obtained by the exhaustive search. These analysis results can provide help in designing a multiuser MMC with optimal resource allocation.