Gradient Descent-Based Parameter Estimation for Ultra Short Baseline

Abstract

This paper proposes a phase-based pseudorange estimation and a gradient descent-based localization algorithm to improve the positioning performance of the ultra short baseline (USBL) system. Traditional USBL systems estimate the pseudorange by cross-correlating the received chirp signal. Still, the limited sampling rate makes it difficult to estimate an accurate pseudorange, resulting in low positioning performance. The proposed method calculates the number of wavelengths of the received chirp signal and detects the initial phase of the signal to estimate a more precise pseudorange than existing methods. Additionally, the localization performance is enhanced using an optimization algorithm based on gradient descent. Through computer simulations and ocean experiments, the proposed method demonstrated up to 29 times better localization performance in computer simulations and up to 107 times better performance in ocean experiments compared to existing methods, verifying that the proposed method outperforms the existing methods.

Keywords

• Gradient descent algorithm
• pseudorange
• source localization
• ultra short baseline (USBL)
• underwater sensor networks