Invariant Extended Kalman Filtering with Partial Orientation Measurement Integration: Theoretical Derivations and Application to Autonomous Surface Vessels

Abstract:Autonomous surface vessels (ASVs) are increasingly vital for marine science, offering robust platforms for underwater mapping and inspection. Accurate state estimation, particularly of vehicle pose, is paramount for precise seafloor mapping, as even small surface deviations can have significant consequences when sensing the seafloor below. To address this challenge, we propose an Invariant Extended Kalman Filter (InEKF) framework designed to integrate partial orientation measurements. While conventional estimation often relies on relative position measurements to fixed landmarks, open ocean ASVs primarily observe a receding horizon. We leverage forward-facing monocular cameras to estimate roll and pitch with respect to this horizon, which provides yaw-ambiguous partial orientation information. To effectively utilize these measurements within the InEKF, we introduce a novel framework for incorporating such partial orientation data. This approach contrasts with traditional InEKF implementations that assume full orientation measurements and is particularly relevant for vehicles operating in a \say{semi-planar} environment, where the attitude is characterized by a dominant yaw rotation with limited roll and pitch variations. This paper details the developed InEKF framework; its integration with horizon-based roll/pitch observations and dual-antenna GPS heading measurements for ASV state estimation; and provides a comparative analysis against the InEKF using full orientation and a Multiplicative EKF (MEKF). Our results demonstrate the efficacy and robustness of the proposed partial orientation measurements for accurate ASV state estimation in open ocean environments.

Originally presented at the 2025 IEEE ICRA Workshop on Field Robotics, extended to cover real world experiments. 13 pages, 8 figures, 3 tables

Robotics (cs.RO)

arXiv:2506.10850 [cs.RO]

(or arXiv:2506.10850v2 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2506.10850 Focus to learn more arXiv-issued DOI via DataCite

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.