Swarm: A decentralized MARL system to locate people:disaster
SWARM is an AI-powered, decentralized search-and-rescue platform that coordinates autonomous drones and rovers to explore disaster zones, ma
Built at NovaStark Hacks 2026
Team The Techtonics: Om Rashiya, 24CS094 SHAH PURVA MANISHKUMAR
Built with React.js, TypeScript, Three.js, Python, PyTorch, MARL, Recurrent PPO, LSTM, ROS 2, Gazebo, SLAM, A* Pathfinding, Node.js.
About the project
SWARM (Search & Rescue Autonomous Multi-Agent System) is an AI-powered, decentralized swarm robotics platform designed to revolutionize disaster response operations. Leveraging Multi-Agent Reinforcement Learning (MARL), the system coordinates autonomous drones and ground rovers to navigate complex, GPS-denied environments such as collapsed buildings, earthquake zones, and hazardous disaster sites. Each robot independently explores its surroundings, builds local maps, and shares critical information through peer-to-peer communication, enabling collaborative navigation, obstacle avoidance, and efficient survivor detection. The platform integrates intelligent route planning, real-time 3D visualization, and AI-driven decision-making to assist rescue teams in identifying survivors and planning safer rescue operations. Currently developed as a simulation-based prototype, SWARM aims to make disaster response faster, smarter, and safer while reducing risks to human rescuers.