Reconfigurability in Wireless Networks: Applications of Machine Learning for User Localization and Intelligent Environment

Abstract

With the rapid development of machine learning (ML) and deep learning (DL) methodologies, the theoretical foundation of leveraging DL in wireless network reconfigurability and channel modeling is studied and summarized. While deep learning based methods have been applied in a few wireless network use cases, there is still much to be explored in many wireless channel modeling scenarios such as predicting channel state information (CSI), and configuring intelligent surface for optimum performance. In this paper, we perform an extensive research on the application of deep learning methods in reconfigurable wireless modeling problems which contains two scenarios. In the first scenario, a user transmitter was moving randomly within a campus area, and at certain spots sending wireless signals that were received by multiple antennas. We constructed an active deep learning architecture to predict user locations from received signals after dimensionality reduction, and analyzed 4 traditional query strategies for active learning to improve the efficiency of utilizing labeled data. We proposed a new location based query strategy that considers both spatial density and model uncertainty when selecting samples to label. We show that the proposed query strategy outperforms all the existing strategies. In the second scenario, a reconfigurable intelligent surface (RIS) containing 4096 tunable cells reflects signals sending from a transmitter to users in an office for better performance. We use the training data of one user's received signals under different configurations to learn the impact behavior of the RIS on the wireless channel. Based on the context and experience from the first scenario, we built a DL neural network that maps RIS configurations to received signal estimations. In a second phase back propagation, the loss function was customized towards our final evaluation formula in order to obtain the optimum configuration array for a user. A further research on the identification of line-of-sight (LOS) and none line-of-sight (NLOS) users has been conducted, which enabled us to prioritize NLOS users over LOS users in our loss function to maximize the final evaluation goal. We built a customized DL pipeline that automatically learns the behavior of RIS on received signals, and generates the optimal RIS configuration array for each of the 50 test users.