Abstract
Fifth generation (5G) networks introduces new paradigm to wireless technology and communications. Millimeter wave is class of Radar technology which has very short wavelength and high frequency bands. Radar system transmit wave signals, which is sometimes reflected by objects in paths. By capturing the reflected signal, a radar system can determine velocity, angle of objects and range. Antennas that are used in 5G technology are smaller than the large antennas used in previous networks. Massive MIMO is used to increase the data rate. In a beamforming technique, the base station continuously calculates the best route for radio waves to reach each mobile device which act as receiver and organize multiple antennas to work together to create beams of millimeter waves to reach at receiving mobile device. In this project, we applied different neural network models over beamformed fingerprints which are Convolution Neural Network, Deep Convolution Gaussian Process, LSTM and Gaussian Process LSTM. Main aim of this project is to minimize the location error almost near to 1 meter in outdoor environment. For that I have used dataset of Millimeter wave published by New York University. Using short sequences with Gaussian Process Long short-term memory neural networks, results shows we can minimize the average error of 1.292 meters in outdoor positioning system.