Data-Driven Iterative Learning Formation Control for Non-affine Multi-Agent Systems Using Integral Sliding Mode

This repository contains MATLAB code and simulation setups for our paper:

📄 Title: Data-driven Iterative Learning Formation Control of Non-affine Multi-agent Systems Using Integral Sliding Mode Method
📅 Published in: 2021 7th International Conference on Control, Instrumentation and Automation (ICCIA)
🔗 DOI: 10.1109/ICCIA52082.2021.9403572

---

🧠 Abstract

The problem of formation control for unknown nonlinear and non-affine multi-agent systems with external perturbations is addressed in this study based on a data-driven robust learning-based algorithm. Firstly, by using the iterative dynamic-linearization concept and pseudo-partial-derivatives, virtual linearized data models are established to represent the agents' unknown dynamics. Subsequently, an iteration-dependent integral sliding variable is considered to design the proposed data-based iterative-learning integral sliding-mode formation control which is solely based upon the input/output information of the agents. Finally, the mathematical analysis proves the stability of the closed loop plant. A demonstrative example is also conducted to verify the proficiency of the designed method.

---

🎯 Overview

This work proposes a robust data-driven formation control strategy for multi-agent systems (MAS) with unknown nonlinear and non-affine dynamics under external disturbances. Key contributions include:

• Virtual linear data models via iterative dynamic-linearization and pseudo-partial-derivatives (PPDs) to capture unknown agent dynamics.
• Iteration-dependent integral sliding variable to design an integral sliding-mode formation controller using only I/O data.
• Robustness to external perturbations without requiring explicit models.
• Stability proof demonstrating convergence of iterative sliding variables and closed-loop stability.
• Demonstrative example validating formation achievement and robustness.

---

🛠 Requirements

• MATLAB R2018b or newer

---

▶️ Usage

1. Clone the repository:

   git clone https://github.com/Babak-Esmaeili/Data-driven-Iterative-learning-Multi-agent-Control.git
   cd Data-driven-Iterative-learning-Multi-agent-Control/Codes

2. Open MATLAB and navigate to the Codes/ folder.
3. Run main.m to reproduce:
   • Agent outputs and formation trajectories at various iterations
   • Formation errors convergence over iterations
   • Control input evolution and robustness analysis

---

📜 License and Contact

This project is licensed under the MIT License – see LICENSE.
For questions or collaboration, contact:

• Babak Esmaeili – esmaeil1@msu.edu

---

📚 Citation

If you use this repository, please cite:

@inproceedings{esmaeili2021iccia,
  title={Data-driven Iterative Learning Formation Control of Non-affine Multi-agent Systems Using Integral Sliding Mode Method},
  author={Esmaeili, Babak and Baradarannia, Mahdi and Salim, Mina},
  booktitle={2021 7th International Conference on Control, Instrumentation and Automation (ICCIA)},
  year={2021},
  organization={IEEE},
  doi={10.1109/ICCIA52082.2021.9403572}
}

---