Recurrent Neural Network Based on DNA Strand Displacement Circuits and Its Application in Location Prediction

With the rapid development of science and technology, various intelligent devices continuously generate large amounts of real-time location data. Efficiently utilizing this data for accurate location prediction has become a critical issue in fields such as intelligent transportation and smart logistics. To realize low-power location prediction, this paper constructs a molecular recurrent neural network (RNN) model based on DNA strand displacement (DSD) technology. The RNN model can process sequence data and accurately predict the next position. Firstly, multiple modules are designed based on DSD circuits, including dual-channel weighted summation module, dual-domain data module and Tanh activation function module. Secondly, a RNN model for processing sequence data is constructed using the above modules. Finally, the constructed RNN model successfully achieves position prediction for multiple inputs and a single output. The robustness and accuracy of the neural network are verified through data experiment. It has been demonstrated that DNA molecules can effectively process complex sequence data. This method holds significant potential in the field of path planning. This paper uses MAE and RMSE to evaluate the experimental data. The results prove that the RNN model constructed in this paper demonstrates strong accuracy and stability.