Agent-based Test Support Systems (AbTSS)

This repository contains the official implementation of the paper:

"TOWARDS AGENT-BASED TEST SUPPORT SYSTEMS: AN UNSUPERVISED ENVIRONMENT DESIGN APPROACH"
by Collins O. Ogbodo, Timothy J. Rogers, Mattia Dal Borgo, and David J. Wagg
Preprint available on arXiv

---

Key Features

• UPOMDP-based Formulation — Underspecified partially observable Markov descision process.
• Dual Curriculum Design — Unspecified enviornment design.
• Parameterised Test Environment Parameterise test environment by frequency range, geometry, boundary condition, simulated damages e.t.c
• Adaptive Sensor Placement Strategy Learn sensor placement across enviornment parameter distribution.
• Information-Theoretic Reward — Maximises determinant of the FIM for informative sensor placement.
• Spatial Correlation-Aware — Rewards spatially well-distributed sensor configurations.
• Case Studies — Applied to a cantilever plate test environment.

---

Enviornment frequency segmentation

Setup

To install the necessary dependencies, run the following commands: Install Ansys Mechanical Student Version

conda create --name dcd python=3.8
conda activate dcd
pip install -r requirements.txt
git clone https://github.com/openai/baselines.git
cd baselines
pip install -e .
cd ..
pip install pyglet==1.5.11

---

Run

python train.py
python eval.py

---

Citation

@article{ogbodo2025towards,
  title={Towards Agent-based Test Support Systems: An Unsupervised Environment Design Approach},
  author={Ogbodo, Collins O and Rogers, Timothy J and Borgo, Mattia Dal and Wagg, David J},
  journal={arXiv preprint arXiv:2508.14135},
  year={2025}
}