Skip to content

Bio-Inspired Underwater Glider Dataset

Overview

Benchmark dataset for dimensionality reduction and optimization of a bio-inspired autonomous underwater glider.

The dataset is designed to support PME-based methods, surrogate modeling, and multi-fidelity optimization workflows.

DOI: https://zenodo.org/records/18936594

Dataset Description

  • Number of configurations: 16,385
  • Design variables: 32
  • Sampling strategy: Sobol sequence
  • Geometry representation: multi-section parametric model
  • Discretization: 784 spatial elements

Data Structure

The dataset follows a structured representation:

  • Geometry (D)
    3 × 784 matrix (spatial coordinates)

  • Design variables (U)
    32-dimensional parameter vector

  • Physics (optional)
    Pressure coefficient (Cp) distribution

  • Outputs

  • Drag
  • Lift

Simulation Setup

  • Flow model: potential flow solver with viscous correction
  • Freestream velocity: 0.25 m/s
  • Angle of attack: 8°
  • Fluid density: 1030 kg/m³

Use Cases

  • Dimensionality reduction (PME, PI-PME, PD-PME)
  • Surrogate-based optimization
  • Multi-fidelity modeling
  • Explainable design-space compression

Files

The dataset includes:

  • database.mat
  • range_design.mat
  • metadata files

Reproducibility

The dataset is fully documented and directly usable within PME-toolkit workflows.