A Model-Based Virtual DVL for Underwater Vehicle Navigation

This paper presents a model-based Virtual Doppler Velocity Log (Virtual DVL) framework for estimating underwater-vehicle linear velocity when direct DVL measurements are unavailable, degraded, or unreliable. The proposed method is built on a physics-based dynamic model of the vehicle and is designed for onboard use with limited computational resources. Starting from the standard six-degree-of-freedom underwater dynamics, the formulation is reduced to the horizontal plane (surge and sway motion are considered), inverted to compute acceleration terms, and integrated in discrete time to reconstruct body-fixed velocity components. The approach is experimentally validated on the reconfigurable RUVIFIST vehicle in hovering configuration by comparing the estimated velocities with measurements acquired from the onboard DVL. Experimental results show that the proposed Virtual DVL is capable of accurately reconstructing the vehicle velocity while maintaining limited drift over medium-range navigation segments, with a mean error of 1.82m. The analysis also highlights reduced estimation accuracy during deceleration phases, suggesting the need for further refinement of hydrodynamic parameter modeling, particularly added-mass effects. Overall, the proposed framework demonstrates the feasibility of using a lightweight model-based Virtual DVL as a robust fallback solution during temporary DVL outages, improving navigation resilience in underwater operations.