Using a generalized regression neural network to improve the accuracy of autonomous navigation in conditions of unstable reception of global positioning system signals

Autonomous navigation is very important in many fields and applications and it specifically depends on global positioning system (GPS) measurements which may not be accessible in some areas. This will directly affect the autonomous navigation and sequentially this will lead to problems according to the function of autonomous navigation. In this research, generalized regression neural network (GRNN) which is a variation to radial basis neural networks, was used to compensate global positioning system (GPS) measurements in case of GPS absences to increase accuracy of autonomous navigation parameters (basically location and velocity) of object. GRNN is integrated with loosely coupled Extended Kalman Filter (EKF). Location, velocity, orientation parameters and biases of sensors are estimated. The evaluation of this methods was conducted using dataset from Internet, two simulations for the GPS measurements outages were made (first outage periods were 35 and 60 seconds) to evaluate the behavior of GRNN, the results shows that using GRNN in GPS absence is effective and robust, it outperformed the only loosely coupled EKF method.