Localization Using Stochastic Gradient Descent Method in a 5G Network

In this paper, we propose an iterative method for smartphone localization in a 5G network. The location estimation accuracy degrades for an inbuilt GPS smartphone in dense environments and indoor scenarios. We propose a combined model that uses the Received Signal Strength (RSS) from macrocells and femtocells thereby increasing the localization accuracy. The location is estimated by optimizing the path loss model for Macro Base Station (MBS) and femtocells using Gradient Descent (GD) and Stochastic Gradient Descent (SGD) methods. Results show that the SGD method outperforms the GD method in terms of computational complexity. Both methods give similar performance for error in distance estimation with an accuracy of 0.89 m. The average localization error with prior knowledge of the location of MBS and femtocell is 2.35 m. Additionally, we derive the textbook derivation of Cramèr Rao Lower Bound (CRLB) to analyze the performance of the estimator. The method successfully localizes smartphones in a dynamic environment, thereby making it suitable for 5G localization.