Decentralized Formation Control for Swarm Quadcopters

Improved Artificial Potential Field (IAPF) + Event-Based Reconfiguration Control (ERC)
A Python simulator of a 5-quadcopter swarm with full 6-DoF dynamics, cascaded PID control, obstacles and wind gusts.

_{ERC swarm shrinks its V-formation, switches to single-file tailgating to pass a 1 m gap, then re-forms.}

---

✨ Highlights

• Hierarchical and decentralized. A swarm-level planner (IAPF / ERC) generates setpoints, and each drone runs its own cascaded PID loop.
• Adaptive reconfiguration. ERC estimates the free width ahead, scales the formation, and switches to tailgating when the gap is too narrow.
• 3D mission extensions. Autonomous takeoff, 3D goal/waypoint navigation, circular-path tracking, and formation orientation along the heading.
• High-fidelity plant. Quaternion 6-DoF model derived with Kane's method (DJI F450 parameters), 2nd-order motor dynamics, and a periodic gust wind model.

🎬 Demo

Scenario	IAPF (static)	ERC (adaptive, proposed)
1 · Basic obstacle avoidance
2 · Narrow gap passage	❌ stalls at the entrance · ▶ YouTube	✅ scale → tailgate → re-form · ▶ YouTube

3 · Narrow gap + wind gust	4 · U-shaped trap (limitation)
▶ V-shape · ▶ Polygon	Local minimum of a purely reactive planner
5 · Waypoint navigation (gust)	6 · Circular path tracking (gust)
▶ YouTube	▶ YouTube

▶ Watch the full simulation playlist on YouTube

🧠 Method at a Glance

System architecture: high-level planner + per-agent PID

High-level planner: state machine, behaviors, event trigger

Behaviors. Every agent sums simple velocity behaviors. ERC blends formation and tailgating with a switching function σᵢ:

$$\mathbf{v}_{ref}^{ERC} = \mathbf{v}^{mig} + \mathbf{v}^{obs} + \mathbf{v}^{col} + \sigma_i,\mathbf{v}^{form} + (1-\sigma_i),\mathbf{v}^{tail}$$

Formation

Migration

Avoidance

Tailgating

Takeoff

V-shape topology

Polygon topology

Free-width estimation we that triggers scaling/tailgating

➡️ Dynamics model, controller gains, full behavior equations and the hardware plan are in docs/TECHNICAL.md.

📊 Key Results

Scenario	Planner	Outcome	Time (s)	Avg. speed (m/s)	RMSEtotal (m)	Φ (order)
1 · Basic obstacles	IAPF	✅	66.790	0.492	0.236	0.892
	ERC	✅	65.745	0.503	0.241	0.890
2 · Narrow gap	IAPF	❌	54.195	0.365	0.525	0.868
	ERC	✅	59.735	0.561	0.692	0.878
3 · Narrow gap + gust	ERC	✅	59.735	0.557	0.694	0.925

IAPF: rigid formation gets stuck at the gap	ERC: formation → tailgating → formation
ERC mode switching and formation scale factor	Control effort with vs. without gust
Waypoint navigation	Circular path tracking

🚀 Quick Start

git clone https://github.com/izmaherdian/multi-agent-sim.git
cd multi-agent-sim
pip install numpy==1.26.4 scipy==1.15.2 sympy==1.13.3 matplotlib==3.9.4
python main.py

Configure the run in agent/config.py:

Setting	Options
CONTROLLER	'erc', 'iapf'
OBSTACLE_SCHEME	'scheme1' basic · 'scheme2' narrow gap · 'scheme3' U-trap · 'NONE'
WIND_TYPE	'NONE', 'FIXED', 'GUST'
PATH_TYPE	'goal', 'multi-goal' (waypoints), 'circular'

At the bottom of main.py, switch the call to main_multi_agent() for the swarm (the default main() runs a single quadcopter). Logs (*.pkl) and videos (*.mp4) are saved to results/data/ and results/videos/ (git-ignored). Post-process them with the scripts in visualization/, e.g. plot_paper_figures.py.

📁 Repository Structure

├── agent/            quadcopter model, cascaded PID, IAPF & ERC planners, config
├── environment/      obstacles, wind model, 3D animation canvas
├── visualization/    analysis & plotting scripts
├── main.py           simulation entry point
├── scratch/          early prototype scripts (not needed to run the simulator)
├── paper/latex/      LaTeX source + figures of the journal paper
└── docs/
    ├── paper/        published paper (PDF) + citation (.ris)
    ├── thesis/       undergraduate thesis report & slides (Bahasa Indonesia)
    ├── media/        GIFs and figures used in this README
    └── TECHNICAL.md  detailed model, controller and algorithm description

📄 Publication & Citation

📑 Read the paper (PDF) · Publisher page · LaTeX source · Thesis report

I. A. Herdian, E. Ekawati, F. Mukhlish, P. Prabaswara, "Decentralized formation control system design for swarm quadcopters using an improved artificial potential field and event-based reconfiguration control," Journal of King Saud University – Engineering Sciences, 38(5), 41 (2026).

@article{Herdian2026,
  author  = {Herdian, Izma Alhazmi and Ekawati, Estiyanti and Mukhlish, Faqihza and Prabaswara, Pramoda},
  title   = {Decentralized formation control system design for swarm quadcopters using an improved artificial potential field and event-based reconfiguration control},
  journal = {Journal of King Saud University -- Engineering Sciences},
  volume  = {38},
  number  = {5},
  pages   = {41},
  year    = {2026},
  doi     = {10.1007/s44444-026-00111-4}
}

A RIS file is available at docs/paper/herdian2026_citation.ris.

📜 License

The source code is released under the MIT License. The paper, thesis and figures in docs/ and paper/ remain under their respective copyrights (the published paper is open access under its publisher's license).

_{Engineering Physics, Institut Teknologi Bandung}