Underwater Robot Localization Using Magnetic Induction: Noise Modeling and Hardware Validation

Abstract

Localization is a fundamental task in many swarm robotic applications, such as foraging and exploration. Magnetic induction communications, which rely on the magnetic component of an antenna's near-field, have gathered interest as means to perform localization in underground and underwater environments. MI signals propagate through lossy environments better than traditional RF signals, and can offer advantages over acoustics. In prior work, we developed a localization method based on MI signals to calculate the range between two moving MI antennas. A core aspect of the method is a particle filter that relies on the received signal strength and speed of the antennas to produce location estimates. In this paper, we first empirically find the amount of noise present in our signal strength measurements in underwater environments. Then we propose a model to capture the impact of the noise on the range calculations and apply it to improve the particle filter's location estimations.