Abstract

Autonomous marine guidance in GPS-denied environments increasingly relies on Earth-based localization using naturally occurring geophysical fields. Among these, the Earth’s magnetic field is particularly promising due to its global presence and long-term stability. However, existing bio-inspired geomagnetic guidance approaches predominantly rely on pre-built magnetic maps, which are impractical for long-range oceanic missions. Drawing inspiration from migratory sea turtles, this paper proposes a GPS-independent homing guidance algorithm based on an Adaptive Biased Random Walk, evaluated at the guidance layer with the control layer idealized. At each step, the vehicle’s heading is sampled from a multimodal distribution of Wrapped Gaussian components. Their means and variances adapt continuously based on magnetic measurements collected along the trajectory. Under a realistic sensor model, simulation results indicate that the vehicle converges to the vicinity of a designated long-range target, relying solely on local magnetic measurements and prior knowledge of the target magnetic signature. No pre-existing geomagnetic map is required. Beyond its engineering relevance, as a separate qualitative observation, the simulated trajectories are broadly consistent with sea turtle migration data, supporting the biological plausibility of the approach.