This paper presents a methodology to study the need and implementation of a GPS-denied navigation system that gives position, velocity and time (PVT) graph. We discuss one such technology that uses an inertial measurement unit (IMU) comprising of accelerometer, gyroscope, magnetometer, and altimeter. We investigate the input and output relationship between a GPS available navigation system (Google Maps) and a GPS-denied navigation system (IMU based AI board). We delineate how to make such a system in a block-wise fashion so that future researchers can get a head start to the area of research that is termed pedestrian dead reckoning (PDR). We show our implementation here which is currently better than the bulk of the research that was found during our time of literature survey. This implementation takes ideas from liquid (NN) machine learning, hybrid convolutional neural network deep learning based online PDR navigation, and generative adversarial network (GAN) based motion transfer learning.
|