Iteratively weighted dynamic modeling of four degrees of freedom motion for marine surface vehicles for k-step ahead prediction

Alper Zihnioğlu*, Melek Ertogan

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


Estimating surge, sway, roll, and yaw motions in marine vehicles are essential for many different reasons, such as maneuvering, dynamic positioning, control, and safety. In this article, an iterative-weighted model is presented for predicting marine vehicle's motions. Surge, sway, roll, and yaw speeds are used as input/output pairs, rudder and throttle references are also used as inputs. By using these input-output relations as training data, an iteratively weighted modeling algorithm is developed. The iteratively weighted model is then used for k-step ahead prediction for different maneuvering scenarios in calm waters. Estimation results are validated with varying scenarios of maneuvering to show the generalization of iteratively progressed model results. The presented work aims to bring a different perspective from a practical use viewpoint of a dynamic model of the platform without knowing the structure of the model, where only needed information are the excitation inputs and required outputs. Therefore, the presented approach has the flexibility to define input and output variables so that it can be configured for the desired application and has the potential to be extended on purpose, increasing the precision and reliability of surface/underwater vehicle's control as well as unmanned vehicles.

Original languageEnglish
Article number110614
JournalOcean Engineering
Publication statusPublished - 15 Feb 2022

Bibliographical note

Publisher Copyright:
© 2022 Elsevier Ltd


  • Parameter estimation
  • Ship dynamic model identification
  • Ship motion prediction
  • Transfer function


Dive into the research topics of 'Iteratively weighted dynamic modeling of four degrees of freedom motion for marine surface vehicles for k-step ahead prediction'. Together they form a unique fingerprint.

Cite this