Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Practical identification approach for the actuation dynamics of autonomous surface vehicles with minimal instrumentation: extended version

Thalia Morel, Luis Orihuela, Christophe Combastel, Guillermo Bejarano

TL;DR

This work addresses practical identification of ASV actuation dynamics and inertia with minimal instrumentation by developing grey-box models that couple propeller actuation to vessel motion. It compares a second-order static propeller model with a first-order dynamic propeller model, both identified from GNSS/AHRS data and PWM signals without accelerometers, and validates them on the Yellowfish catamaran. The results show robust, high-fidelity predictions (R^2 > 0.94, MAE in the 0.01–0.03 m/s range) for surge, sway, and yaw, demonstrating the feasibility of accurate autonomous navigation and control with limited sensors. The study provides detailed identification procedures, parameter tables, and publicly available data to enable replication and further development of low-instrumentation ASV identification methods.

Abstract

A practical method for identifying the propeller model and inertia matrix of a marine Autonomous Surface Vehicle (ASV) is proposed in this work. Special attention is paid to limiting the instrumentation requirements. Based on a generic grey-box dynamic modelling addressing the considered catamaran-shaped ASV architecture, the static/dynamic behaviour of both propellers and the vessel dynamic are jointly estimated using the sole measurements of position, heading, and propellers pulse width modulation (PWM) signals. No accelerometer is required. Two distinct grey-box configurations involving either a static polynomial or a dynamic modelling of each propeller are proposed and compared. The resulting ASV identification methodology is shown to provide insight into the whole vessel inertial characteristics, which are key enablers in the development of autonomous navigation and control systems. Model validation was performed using data collected from the reported experiments. Model prediction errors related to both linear velocities and yaw rate are evaluated and compared based on given metrics. The results underscore the robustness and accuracy of the identified models in capturing the essential dynamics of the ASV, with a determination coefficient that consistently exceeds 0.94 for all estimated velocities.

Practical identification approach for the actuation dynamics of autonomous surface vehicles with minimal instrumentation: extended version

TL;DR

This work addresses practical identification of ASV actuation dynamics and inertia with minimal instrumentation by developing grey-box models that couple propeller actuation to vessel motion. It compares a second-order static propeller model with a first-order dynamic propeller model, both identified from GNSS/AHRS data and PWM signals without accelerometers, and validates them on the Yellowfish catamaran. The results show robust, high-fidelity predictions (R^2 > 0.94, MAE in the 0.01–0.03 m/s range) for surge, sway, and yaw, demonstrating the feasibility of accurate autonomous navigation and control with limited sensors. The study provides detailed identification procedures, parameter tables, and publicly available data to enable replication and further development of low-instrumentation ASV identification methods.

Abstract

A practical method for identifying the propeller model and inertia matrix of a marine Autonomous Surface Vehicle (ASV) is proposed in this work. Special attention is paid to limiting the instrumentation requirements. Based on a generic grey-box dynamic modelling addressing the considered catamaran-shaped ASV architecture, the static/dynamic behaviour of both propellers and the vessel dynamic are jointly estimated using the sole measurements of position, heading, and propellers pulse width modulation (PWM) signals. No accelerometer is required. Two distinct grey-box configurations involving either a static polynomial or a dynamic modelling of each propeller are proposed and compared. The resulting ASV identification methodology is shown to provide insight into the whole vessel inertial characteristics, which are key enablers in the development of autonomous navigation and control systems. Model validation was performed using data collected from the reported experiments. Model prediction errors related to both linear velocities and yaw rate are evaluated and compared based on given metrics. The results underscore the robustness and accuracy of the identified models in capturing the essential dynamics of the ASV, with a determination coefficient that consistently exceeds 0.94 for all estimated velocities.
Paper Structure (20 sections, 3 theorems, 48 equations, 10 figures, 11 tables)

This paper contains 20 sections, 3 theorems, 48 equations, 10 figures, 11 tables.

Table of Contents

  1. Introduction
  2. Problem Statement
  3. Experimental Platform, Instrumentation and Available Data
  4. Modelling
  5. ASV body
  6. Actuation System
  7. Second-Order Static Model
  8. First-Order Dynamic Model
  9. Identification Procedure
  10. Data Preparation
  11. Input-Gain Identification for Static Propeller Modelling
  12. Input-Gain Identification for Dynamic Propeller Modelling
  13. Identification Experiments, Results and Discussion
  14. Experimental Setup and Identification Metrics
  15. Identification Results for Static Propeller Modelling
  16. ...and 5 more sections

Key Result

Proposition 1

Provided that $\bar{\delta}(k)$ and $\Delta\delta(k)$ are known for every $k$, and the velocities $u(k),v(k),r(k)$ can be obtained from the position data as explained in Section sec:data_prep, the evolution of the surge at instant $k$ for a quadratic static model for the propeller and operating in t

Figures (10)

  • Figure 1: Reference Frames of the Vessel
  • Figure 2: The Yellowfish ASV Experimental Platform
  • Figure 3: Block Diagram of the ASV, Highlighting the Actuation System and the ASV Body
  • Figure 4: The Experiment Field and ASV Path
  • Figure 5: Surge Velocity $u$ and Estimated Velocity $\hat{u}$ in the Validation Subset Using Second-Order Static Model
  • ...and 5 more figures

Theorems & Definitions (12)

  • Remark 1
  • Proposition 1
  • proof
  • Remark 2
  • Proposition 2
  • proof
  • Remark 3
  • Proposition 3
  • proof
  • Remark 4
  • ...and 2 more