A Robust Disturbance Compensation Method for Integrated Navigation Systems based on Gaussian Process Regression

The Doppler velocity log (DVL), a velocity measurement instrument used in navigation systems, functions in dual operational states: bottom-tracking (BT-DVL) and water-tracking (WT-DVL), with each state determining the velocity relative to the seabed and water flow, respectively. However, in deep-sea environments, DVL may lose bottom-tracking capability and include disturbance from ocean currents. To address this challenge, this paper proposes a robust disturbance compensation method for INS/DVL integrated navigation systems based on Gaussian process regression (GPR). Based on the probabilistic modeling capabilities of GPR, this method accurately estimates the complex spatiotemporal dynamics of ocean currents, thereby improving navigation precision in scenarios where bottom-tracking is unavailable. Experimental results verified the effectiveness of the algorithm.