Robust Positioning Method for Berthing Support Using Fusion of Tightly Coupled GNSS Carrier Phase with IMU and Visual Odometry

Hiraku Nakamura, Tatsuya Sonobe, Hiroyuki Toda, Naomi Fujisawa, Takafumi Taketomi, Alexander Plopski, Christian Sandor, Hirokazu KatoJournal of the Institute of Positioning Navigation and Timing of Japan 2019 · DOI

Abstract

船舶の着桟支援を想定したロバストな高精度測位手法を開発した.本手法は,自船に固定したカメラで撮影した動画像データ,複数のGNSSアンテナから受信した民生用L1信号データ,及び,6軸IMUデータの融合により実現する.動画像データの視覚オドメトリを用いて,港湾構造物の3次元構造を復元しつつ,6軸IMUデータとGNSS衛星単位の搬送波位相の時空間変化量とのタイトカップリングによって高精度に自船の軌跡を推定する新しいオドメトリ技術を構築した.さらに,クロックを共有した複数のGNSS受信機のクロックドリフト推定を共通化することで,約10cmの高精度な軌跡精度を維持するための衛星数の条件が緩和できたことを確認した.

Introduction

Automating ship berthing is a critical challenge for Maritime Autonomous Surface Ships (MASS). The Japanese Ministry of Land, Infrastructure, Transport and Tourism specifies a precision requirement of 10 cm for safe berthing. Achieving this in port environments is difficult; standard GNSS lacks precision, while high-precision RTK-GNSS requires shore-side infrastructure that is often unavailable. Onboard sensors like LIDAR and stereo cameras face limitations in range and physical baseline requirements. This paper proposes a robust, infrastructure-free positioning system that fuses monocular visual odometry (VO), a 6-axis IMU, and tightly coupled GNSS carrier phase data to meet these stringent requirements.

Methodology: Tightly Coupled Sensor Fusion

The core of the system is an Extended Kalman Filter (EKF) that integrates three distinct data streams to estimate the ship's trajectory and the 3D structure of the port.

System architecture of the sensor fusion algorithm integrating VO, GNSS, and IMU data.
  1. Monocular Visual Odometry (VO): This component recovers the 3D structure of the pier and estimates relative motion. While VO is high-resolution, it typically suffers from scale ambiguity and long-term drift.
  2. 6-axis IMU: Provides high-frequency inertial data to bridge gaps between other sensor updates.
  3. Tightly Coupled (TC) GNSS: Instead of using processed velocity, this system utilizes raw carrier phase time-space variations. This "anchors" the VO, providing the absolute scale and preventing trajectory drift.

The "Common Clock" Innovation

A key feature of this system is the use of four GNSS antennas sharing a single hardware clock. By treating clock drift as a common state variable in the EKF, the number of mathematical unknowns is reduced. This allows the system to maintain decimeter-level accuracy even in "signal-masked" environments (e.g., under gantry cranes) where only 2 or 3 satellites are visible—a scenario where traditional GNSS methods would fail.

Experimental Evaluation

To validate the system, sea trials were conducted at the Nishinomiya Marina in Hyogo Prefecture, Japan. The experimental setup involved a small test boat equipped with a prototype sensor suite. High-precision RTK-GNSS served as the ground truth for performance evaluation.

The prototype sensor rig, including the camera and GNSS antennas, mounted on the test vessel.

Key Results

  • Precision: In optimal conditions, the system achieved a horizontal trajectory error of 4.0 cm (standard deviation), comfortably meeting the 10 cm target.
  • Robustness: When visibility was restricted to just 3 satellites, the tightly coupled method maintained an error of ~12 cm. In contrast, a loosely coupled approach saw errors exceed 1.5 meters.
  • Clock Sharing Efficacy: In a worst-case simulation where four antennas each saw a different single satellite, the common clock estimation kept the error at 6.3 cm, whereas independent clock estimation resulted in a 1.44-meter error.

Conclusion

The proposed fusion method provides a robust, high-precision positioning solution for ship berthing without requiring shore-side infrastructure. By leveraging tightly coupled GNSS carrier phases and common clock drift estimation, the system remains operational in challenging port environments. Future work will focus on improving reliability in adverse weather conditions (fog, rain) by incorporating additional sensors like radar.

Reference: Nakamura, H., et al. (2019). Robust Positioning Method for Berthing Support Using Fusion of Tightly Coupled GNSS Carrier Phase with IMU and Visual Odometry. Journal of the Japan Institute of Navigation, Vol. 10, No. 1.

Summary updated

Abstract

船舶の着桟支援を想定したロバストな高精度測位手法を開発した.本手法は,自船に固定したカメラで撮影した動画像データ,複数のGNSSアンテナから受信した民生用L1信号データ,及び,6軸IMUデータの融合により実現する.動画像データの視覚オドメトリを用いて,港湾構造物の3次元構造を復元しつつ,6軸IMUデータとGNSS衛星単位の搬送波位相の時空間変化量とのタイトカップリングによって高精度に自船の軌跡を推定する新しいオドメトリ技術を構築した.さらに,クロックを共有した複数のGNSS受信機のクロックドリフト推定を共通化することで,約10cmの高精度な軌跡精度を維持するための衛星数の条件が緩和できたことを確認した.
BibTeX
@article{nakamura2019robust,
  author = {Hiraku Nakamura and Tatsuya Sonobe and Hiroyuki Toda and Naomi Fujisawa and Takafumi Taketomi and Alexander Plopski and Christian Sandor and Hirokazu Kato},
  title = {{Robust Positioning Method for Berthing Support Using Fusion of Tightly Coupled GNSS Carrier Phase with IMU and Visual Odometry}},
  journal = {Journal of the Institute of Positioning Navigation and Timing of Japan},
  year = {2019},
  doi = {10.5266/ipntj.10.1},
}
Figure