Optimized Position Estimation in Multipath Environments using Machine Learning

GNSS receivers in urban areas suffer from degraded positioning accuracy due to reflected signals. When a receiver is moving, the same satellite signal could continue to change its status. A signal status could be Line-Of-sight (LOS), multipath, non-LOS (NLOS), or invisible where the LOS is completely blocked. The change in a signal status presents an extra challenge when developing a multipath mitigation algorithm. This is because an efficient multipath mitigation algorithm should be able to detect changes in a signal status and adjust its performance accordingly. This paper proposes two new algorithms that can work in urban areas, with a moving receiver, to enhance the accuracy of position estimation. The first algorithm is called Optimized Position Estimation (OPE). The OPE algorithm estimates the most likely sequences of positions on a map, where each sequence forms a path. It then finds the path with the optimal weight. Three functions are used in the computation of the weight, including a function based on the probability of transitions between positions. The computation of the probability of transition is done using the second proposed algorithm, which is called Intelligent signal Status Estimation (ISE). The ISE algorithm utilizes a Self-Organizing Map (SOM) neural network, which is a machine learning algorithm, to estimate the probability of a change in a received signal status based on predictions of signal reflections from the surrounding environment. Where, a signal status is changed on the appearance or disappearance of the LOS signal or multipath signals. The features of the SOM neural network are extracted from a tracking module. A conventional SOMnetwork uses unsupervised learning and clusters inputs based on their features. The proposed SOM network uses a supervised learning approach to give a probability to a change in a signal status. The proposed algorithms are tested using GPS C/A signals with a moving receiver in an urban area. The tested scenario has over 50 changes in the satellites signal status. The results indicate that there is an accuracy enhancement of up to 96% over the accuracy of a conventional navigation algorithm.