Swarm Simulation Framework

A flexible, no-code swarm robotics simulation framework built in Unity, designed to simplify multi-agent modeling and testing of nature-inspired algorithms for search and rescue applications.

Overview

This framework addresses a critical gap in swarm robotics research: the time-consuming process of modeling simulations from scratch. By providing reusable, modular components and a visual node-based workflow, researchers can rapidly prototype and test swarm algorithms without extensive coding.

ℹ️ Information

This project is the core simulation framework, but does not contain stigmergy simulation!

Stigmergy simulation was explored in another minimal repository, on a similar Unity environment.

Link: Github Repository for Stigmergy Testing

Key Features

• No-Code Development: Visual block-based workflow using Unity's Bolt toolkit
• Modular Architecture: Standalone units for agent modeling, map generation, and simulation management
• Nature-Inspired Algorithms: Built-in implementations of flocking, cohesion, alignment, and avoidance behaviors
• Procedural Map Generation: Automated terrain generation with customizable obstacles
• Search and Rescue Scenarios: Designed for target localization and pathfinding experiments
• Scalable: Simulate hundreds of agents with efficient computational performance
• 3D Support: Full 3D environment simulation for realistic testing

Repository Structure

Swarm-Simulation-master/
├── Unity/
│   ├── Flocking Algorithm Sandbox/    # Main flocking behavior implementation
│   ├── Stigmergy Sandbox/             # Stigmergy-based communication experiments
│   └── TerrainDemo_v1/                # 3D terrain generation demo
└── Netlogo/
    └── SingleDronePheromoneNoEnvironment.nlogo  # NetLogo prototype

Framework Components

1. Single Agent Modeling

Define individual agent behaviors including:

• Cohesion: Agents cluster together at an average position
• Alignment: Agents steer towards the average direction of neighbors
• Avoidance: Collision avoidance with obstacles and other agents

2. Collective Agent Modeling

Configure swarm-level behaviors:

• Flock size and density control
• Peer-to-peer communication
• Stigmergy-oriented pathfinding
• Emergent behavior observation

3. Procedural Map Generation

• Randomized bitmap generation with configurable seed values
• Smoothing algorithms for organic obstacle placement
• Adjustable map scale and obstacle density
• Reproducible terrains for consistent testing

4. Target Placement

• Custom target positioning for search tasks
• Configurable start locations for agent deployment
• Support for multiple targets and search patterns

5. Simulation Management

• Configurable iteration counts
• Time limits per iteration
• JSON-based result output for analysis
• Performance metrics (collision avoidance, flock cohesion)

Getting Started

Prerequisites

• Unity 2019.4 LTS or later
• Visual Studio or VS Code (for optional code modifications)
• Basic understanding of swarm robotics concepts

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/Swarm-Simulation.git

2. Open Unity Hub and add the project:

   • Navigate to Unity/Flocking Algorithm Sandbox/ or Unity/Stigmergy Sandbox/
   • Open with Unity 2019.4 or later

3. Open the sample scene:

   • Go to Assets/Scenes/SampleScene.unity

Quick Start Example

The framework uses a visual node-based interface. To set up a basic simulation:

1. Generate Map: Set map scale, seed value, and obstacle percentage
2. Define Agent Behaviors: Configure cohesion, alignment, and avoidance parameters
3. Set Agent Count: Specify flock size and maximum cluster size
4. Place Agents: Define home location and spawn radius
5. Add Targets: Set target count and coordinates
6. Run Simulation: Configure iterations and time limits

Example parameters for target localization:

• Map Scale: 1 unit
• Obstacle Density: 35%
• Flock Size: 100 agents
• Cohesion Radius: 1.5 units
• Alignment Radius: 1.5 units
• Avoidance Radius: 0.5 units

Nature-Inspired Algorithms

Flocking Behavior

The framework implements Reynolds' flocking algorithm with three core behaviors:

Flocking = w1 × Cohesion + w2 × Alignment + w3 × Avoidance

Where w1, w2, w3 are adjustable weights for behavior priority.

Stigmergy

Indirect communication through environmental modifications (pheromone trails), inspired by ant colony behavior. Implemented in the Stigmergy Sandbox project.

Simulation Results

The framework generates JSON-formatted results including:

• Collision avoidance metrics per agent
• Time spent in flock formations (as percentage)
• Simulation parameters for reproducibility
• Target localization success rates

Results can be visualized using MATLAB or other analysis tools.

Research Applications

This framework is ideal for:

• Testing swarm intelligence algorithms
• Search and rescue mission simulations
• Multi-agent coordination research
• Emergent behavior studies
• Algorithm performance benchmarking

Comparison with Existing Tools

Feature	This Framework	Stage	Gazebo	Unity (Raw)	NetLogo
No-Code Interface	✅	❌	❌	❌	⚠️
3D Support	✅	❌	✅	✅	✅
Large Agent Count	✅	⚠️	❌	✅	✅
Rapid Prototyping	✅	❌	❌	❌	⚠️
Reusable Components	✅	❌	❌	❌	⚠️

Limitations

• Not all nature-inspired algorithms are fully integrated
• No real-time scenario comparison yet
• Requires Unity installation and basic familiarity
• Performance depends on hardware for large-scale simulations

Future Work

• Agent trajectory tracking and memory management
• Self-learning capabilities using reinforcement learning
• Additional nature-inspired algorithms (particle swarm, genetic algorithms)
• Real-world robot integration and testing
• Performance optimization for larger swarms (1000+ agents)

Technical Details

Built With

• Unity Engine: Game engine and simulation environment
• Bolt Visual Scripting: No-code node-based development
• C#: Core framework components
• TextMesh Pro: UI elements
• NetLogo: Initial prototyping (included for reference)

File Formats

• .unity - Unity scene files
• .cs - C# behavior scripts
• .prefab - Reusable agent prefabs
• .asset - Behavior and filter configurations
• .json - Simulation results

Documentation

For detailed methodology and research findings, please refer to the accompanying thesis document which covers:

• Comprehensive literature review
• Framework architecture and design decisions
• Implementation details and algorithms
• Experimental results and analysis
• Comparisons with existing platforms

Contributing

Contributions are welcome! Areas for improvement:

• Additional agent behaviors and algorithms
• Performance optimizations
• UI/UX enhancements
• Documentation improvements
• Example scenarios and tutorials

License

This project was developed as part of academic research at the Military Institute of Science and Technology.

Citation

If you use this framework in your research, please cite:

[Swarm Simulation Framework]
Military Institute of Science and Technology
Department of Computer Science and Engineering
Supervisor: Dr. Nusrat Sharmin

Acknowledgments

• Dr. Nusrat Sharmin (Supervisor) - Military Institute of Science and Technology
• Md Shadman Aadeeb - Military Institute of Science and Technology
• Department of Computer Science and Engineering, MIST

Contact

For questions, suggestions, or collaboration opportunities, please open an issue on GitHub.

---

Note: This framework represents a proof-of-concept for simplifying swarm robotics research. While functional, it is intended for academic and research purposes.