GPS/INS Integration During GPS Outages Using Machine Learning Augmented with Kalman Filter

Abstract

Nowadays a tremendous progress has been witnessed in Global Positioning System (GPS) and Inertial Navigation System (INS). The Global Positioning System provides information as long as there is an unobstructed line of sight and it suffers from multipath effect. To enhance the performance of an integrated Global Positioning System and Inertial Navigation System (GPS/INS) during GPS outages, a novel hybrid fusion algorithm is proposed to provide a pseudo position information to assist the integrated navigation system. A new model that directly relates the velocity, angular rate and specific force of INS to the increments of the GPS position is established. Combined with a Kalman filter the hybrid system is able to predict and estimate a pseud GPS position when GPS signal is unavailable. Field test data are collected to experimentally evaluate the proposed model. In this paper, the obtained GPS/INS datasets are pre-processed and semi-supervised machine learning technique has been used. These datasets are then passed into Kalman filtering for the estimation/prediction of GPS positions which were lost due to GPS outages. Hence, to bridge out the gaps of GPS outages Kalman Filter plays a major role in prediction. The comparative results of Kaman filter and extended Kalman filter are computed. The simulation results show that the GPS positions have been predicted taking into account some factors/measurements of a vehicle, the trajectory of the vehicle, the entire simulation was done using Anaconda (Jupyter Notebook).