🧠 About the Challenge

Welcome to the GhostSwarm Hackathon, a cutting-edge robotics and AI competition where the boundaries of autonomous systems, multi-agent collaboration, and ROS 2 engineering are pushed to their limits.

In this challenge, participants will design, simulate, or prototype heterogeneous robotic agents (ground and aerial) that operate in adversarial or denied environments—think GPS-jammed zones, unstable terrain, or underground tunnels. Teams must implement distributed intelligence using ROS 2 to enable collaboration without centralized control, while navigating communication delays, partial observability, and deceptive environments.

This hackathon isn’t just about building robots. It’s about solving some of the toughest technical problems in multi-agent coordination, shared autonomy, and resilient perception and planning—the kind of work that could shape the future of defense, disaster response, and planetary exploration.

🚀 Get Started

Ready to take on the GhostSwarm challenge? Here's how to begin:

  1. Form Your Team
    Solo or up to 4 members. Choose collaborators with strengths in robotics, ML, AI, systems, or simulation.

  2. Pick Your Domain

    • Simulation Track: Build and test your system in Gazebo or Webots

    • Hardware Track: Use real-world UGVs/UAVs (TurtleBot, Crazyflie, etc.)

  3. Read the Technical Brief
    Download the [Challenge Kit], which includes the task objectives, constraints, ROS 2 node architecture template, and evaluation criteria.

  4. Build with ROS 2
    Use ROS 2 (Galactic/Humble), open-source tools, and any ML models you prefer. Focus on communication robustness, decentralized autonomy, and agent diversity.

  5. Submit Your Project
    Final deliverables include:

    • Code (GitHub repo)

    • 3-min demo video

    • Architecture diagram

    • Write-up/blog or whitepaper

  6. Compete for Prizes
    Winners will be judged on innovation, autonomy, robustness, and real-world potential. Prizes include mentorship, publication opportunities, and hardware kits.

Requirements

🛠️ What to Build

Your mission is to build a multi-agent robotic system that operates in a denied or adversarial environment using ROS 2 as the core architecture.

You can choose one or both tracks:

🔹 Simulation Track

  • Use Gazebo, Ignition, or Webots to simulate:

    • Aerial (UAV) + ground (UGV) robotic agents

    • Partial observability and unreliable communication

    • Real-time environment exploration, coordination, and decision-making

🔹 Hardware Track (Optional for advanced teams)

  • Use physical robots (TurtleBot, Crazyflie, custom UGVs/UAVs)

  • Demonstrate real-world deployment in constrained or cluttered environments

  • Agents must perform search, mapping, coordination, or delivery tasks under uncertainty

🔧 Core System Requirements

  • ROS 2-based multi-node system (Galactic, Humble, or Iron)

  • Heterogeneous agents (different roles/sensors/abilities)

  • Decentralized or resilient coordination (no central brain)

  • Environmental awareness (SLAM, object detection, or localization)

  • Optional: Incorporate AI/ML (e.g., RL, YOLO, segmentation) for perception or decision-making

📦 What to Submit

Your final submission should clearly showcase your system’s capabilities, design, and impact.

âś… Required Deliverables:

  1. GitHub Repository

    • ROS 2 workspace with structured packages

    • Launch files, simulation worlds, config files

    • Clear README with setup instructions

  2. Demo Video (max 3 minutes)

    • Show system in action (simulated or physical)

    • Highlight coordination, obstacle handling, and autonomy

  3. System Architecture Diagram

    • Illustrate robot roles, communication, ROS 2 nodes, and data flow

  4. Technical Write-Up or Blog (1–3 pages or 1000–1500 words)

    • Describe your approach, key challenges, and innovations

    • Include any ML models used, strategies for coordination, and system limitations

🏆 Bonus (for extra recognition):

  • Submit a research-style paper (PDF or arXiv-ready)

  • Publish your write-up on Medium or DevPost

  • Include benchmarks, maps, or visualizations of multi-agent behavior

Hackathon Sponsors

Prizes

$700 in prizes
Winner
$700 in cash
1 winner

A DIGITAL CERTIFICATE

Devpost Achievements

Submitting to this hackathon could earn you:

Judges

Joushua Jothi Rajan

Judging Criteria

  • Technical Complexity
    Depth of ROS 2 implementation, use of advanced algorithms, system architecture, and performance under constraints.
  • Autonomy & Coordination
    How well agents behave independently and work as a team in dynamic environments. Decentralized or resilient behaviors are highly valued.
  • Innovation & Creativity
    Unique approaches to solving the problem. Novel AI/ML integrations, swarm behaviors, or problem framing.

Questions? Email the hackathon manager

Tell your friends

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.