Machine Learning-Based Screening and Measurement to Measurement Association for Navigation in GNSS Spoofing Environment

Global navigation satellite system (GNSS) provides reliable positioning across the globe. However, GNSS is vulnerable to deliberate interference problem like spoofing, which can cause fake navigation. This article proposes navigation in a GNSS spoofing environment by taking the received power, correlation distortion function, and pseudorange measurement observation space into account. In the proposed approach, both actual and interference measurements are considered a set. Machine learning screens the authentic measurements from the accessible set using parameters such as received power and correlation function distortion. To maintain the track and navigate the GNSS’s time-varying kinematics, we used a combination of the gating technique within the Kalman filter framework and logic-based track management. The machine learning classifiers like support vector machines (SVMs), neural networks (NNs), ensemble, nearest neighbor, and decision trees are explored, and we observe that linear SVM and NN provide a test accuracy of 98.20%. A time-varying position-pull off strategy is considered, and the metrics like position RMSE and track failure are compared with the conventional M-best algorithm. The results show that for four authentic measurements and spoof injections, there are only a few track failures. In contrast, even with an increase in spoof injections, track failures are zero in the case of six authentic measurements.