Impact Assessment of Efficient Denoising Techniques in AI-Based Low-Cost INS/GPS Integration During Blockage of GPS Satellites

Low-cost micro-electromechanical systems (MEMS)-based inertial navigation system (INS) sensors present long-term drift and different stochastic errors over time, which, in turn, tends to degrade the positioning accuracy of the artificial intelligence (AI)-based GPS/INS navigation systems during GPS outages. We address this issue by analyzing and modeling various stochastic errors of these sensors and our efficient noise reduction method can improve the accuracy of such navigation systems. In the present work, we propose our efficient adaptive noise reduction method based on the Allan variance (AV) analysis technique in order to choose more reliable, appropriate, and optimum parameters of the noise reduction method, which can help remove stochastic noises efficiently while preserving vehicles motion information. In our low-cost AI-based GPS/INS navigation system, more accurate and smoother INS sensor measurements with less complexity, as an input of intelligence structure, can provide better training and less estimation error and better generalization ability in the uncertainty-oriented environment. Eventually, in order to assess the effect of the proposed noise reduction method based on the AV method on the low-cost AI-based GPS/INS navigation systems under the GPS signal blockages, two different candidates from AI-based models are selected according to their ability in modeling random noise inputs. Then, their performance with and without the proposed noise reduction method is evaluated. The obtained results indicate that the proposed noise reduction method significantly improves positioning accuracy in AI-based low-cost GPS/INS integrated systems under GPS signal blockages using real data collected even with a high-speed test vehicle.