FrontierMesh

FrontierMesh is an early-stage Java/libGDX simulation sandbox for experimenting with multi-agent movement, local coordination, and decentralized exploration strategies.

The current version focuses on the motion-simulation layer: simple agents move around a 2D world, follow a clicked target, avoid each other, and steer around rectangular obstacles.

The long-term goal is to develop FrontierMesh into a modular simulation platform for studying decentralized multi-agent exploration under limited sensing and communication constraints, with possible applications to IoT/edge, UAV-inspired, and search-and-rescue-style scenarios.

Current features

• libGDX desktop application
• fixed-size 2D world with viewport scaling
• multiple moving drone-like agents
• click-to-target movement
• local separation between agents
• rectangular obstacle avoidance
• basic debug rendering
• Spotless Java formatting

Research direction

The planned research direction is to use FrontierMesh as a platform for comparing decentralized multi-agent exploration strategies in communication-constrained environments.

Possible strategies to implement and evaluate include:

• random or heuristic exploration baselines
• frontier-based exploration
• communication-aware exploration
• relay-aware coordination for maintaining network connectivity
• optional lightweight learning-based policies

Potential evaluation metrics include:

• explored area over time
• time to reach a target coverage level
• fraction of agents connected to a base station
• number and duration of communication breaks
• total movement cost
• robustness under different obstacle densities, swarm sizes, and communication ranges

Tech stack

• Java 25
• Gradle
• libGDX
• LWJGL3
• Spotless

Project structure

core/       Main simulation and rendering code
lwjgl3/     Desktop launcher
assets/     Runtime assets

Running

From the project root:

./gradlew lwjgl3:run

On Windows:

.\gradlew.bat lwjgl3:run

Building

./gradlew build

Formatting

Check formatting:

./gradlew spotlessCheck

Apply formatting:

./gradlew spotlessApply

Recommended before committing:

./gradlew spotlessApply build

Controls

• Left mouse button: move swarm target

Roadmap

Near-term:

• pause, reset, and debug controls
• cleaner steering-policy abstraction
• more obstacle layouts
• fog-of-war exploration
• communication links between agents
• basic experiment metrics

Longer-term:

• modular exploration policies
• communication graph and base-station connectivity
• relay-aware coordination strategy
• headless experiment runner
• CSV/JSON export for evaluation results

License

Apache License 2.0