Counter Adversial GPS Jamming

Military and commercial missions rely heavily on the Global Positioning System (GPS), but GPS signals, due to their weak intensity and their need to travel considerable distances, are vulnerable to external interference such as jamming and spoofing. Adversarial jamming can be strong in that its amplitude can be orders of magnitude larger than that of the GPS signal, and they share the same spectral band. Various methods based mostly on traditional signal-processing techniques, such as filtering or wavelet transform, were proposed to mitigate GPS jamming to a certain extent. Often, the effectiveness of a specific method depends on the type of jamming signals in terms of their waveform, amplitude, and frequency range. GPS antijamming is essentially a nonlinear signal prediction and classification problem. For nonlinear signal prediction, a possible solution is exploiting modern machine learning, where a suitable neural-network architecture trained with features of the desired signals views the jamming as undesired and removes them. However, the inevitable noise represents a difficulty when employing machine learning to predict the GPS waveform and navigation signal, which can significantly degrade the antijamming capability. In nonlinear dynamical systems, the counterintuitive phenomenon of stochastic resonance (references 1 & 3) has been well established, where the signal-to-noise ratio for a weak signal can be enhanced by tuning the noise amplitude. This opens the possibility of enhancing the GPS signal through the stochastic-resonance mechanism.