ResQNet

ResQNet - Agentic AI Disaster Swarm Network

𝗜𝗻𝘀𝗽𝗶𝗿𝗮𝘁𝗶𝗼𝗻

India witnesses frequent floods, cyclones, and landslides that claim thousands of lives each year. The tragedy is rarely due to the lack of resources, but due to poor coordination between agencies, scattered data sources, and language barriers. We were inspired by the idea of a swarm in nature - where ants or bees coordinate seamlessly without a central commander. If nature solves survival problems through distributed intelligence, why can’t disaster relief work the same way? That sparked the creation of our Agentic AI Disaster Swarm Network.

𝗪𝗵𝗮𝘁 𝗶𝘁 𝗱𝗼𝗲𝘀

Our project, ResQNet, is a real-time disaster response coordination system where autonomous AI agents (Drone, Boat, Truck, Relief Center, Communication) collaborate dynamically to:

Fuse multi-source data (weather APIs, NDMA bulletins, social media SOS, drone feeds).
Create disaster heatmaps with high-risk zones.
Allow rescue units to negotiate actions with each other (e.g., drone detects victims → alerts boat → truck reroutes supplies).
Deliver multilingual alerts to responders and citizens in regional languages.

The end result: a 'living swarm brain' that reduces chaos, optimizes resource allocation, and cuts disaster response time by half.

𝗛𝗼𝘄 𝘄𝗲 𝗯𝘂𝗶𝗹𝘁 𝗶𝘁

𝗙𝗿𝗼𝗻𝘁𝗲𝗻𝗱: A React.js dashboard visualizing agents on a map.
𝗕𝗮𝗰𝗸𝗲𝗻𝗱:Node.js server integrated with IBM Agent Development Kit (ADK) to create agentic workflows.
𝗔𝗜 𝗠𝗼𝗱𝗲𝗹𝘀: IBM Granite models for summarization, translation, and multilingual communication.
𝗗𝗮𝘁𝗮 𝗦𝗼𝘂𝗿𝗰𝗲𝘀: Mock APIs (synthetic SOS posts, weather data, drone feeds) to simulate real disaster scenarios.
𝗗𝗲𝗺𝗼 𝗙𝗹𝗼𝘄: Drone agent detects victims → Boat agent reroutes → Truck reallocates food → Communication agent broadcasts updates in English + regional languages.

𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝘄𝗲 𝗿𝗮𝗻 𝗶𝗻𝘁𝗼

𝗗𝗮𝘁𝗮 𝗿𝗲𝗮𝗹𝗶𝘀𝗺: Disaster datasets are often scattered; we had to generate synthetic SOS + relief data to simulate a crisis.
𝗔𝗴𝗲𝗻𝘁 𝗻𝗲𝗴𝗼𝘁𝗶𝗮𝘁𝗶𝗼𝗻: Designing meaningful interactions between agents (without them becoming redundant or chaotic) required fine-tuning logic flows.
𝗧𝗶𝗺𝗲 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀: Building a system that looks futuristic but is still feasible in 12 hours pushed us to focus on simulation + visualization.
𝗠𝘂𝗹𝘁𝗶𝗹𝗶𝗻𝗴𝘂𝗮𝗹 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻: Ensuring alerts made sense in Indian regional languages while keeping them concise.

𝗔𝗰𝗰𝗼𝗺𝗽𝗹𝗶𝘀𝗵𝗺𝗲𝗻𝘁𝘀 𝘁𝗵𝗮𝘁 𝘄𝗲'𝗿𝗲 𝗽𝗿𝗼𝘂𝗱 𝗼𝗳

First-ever Agentic AI Swarm demo that showcases autonomous negotiation between AI-driven rescue units.
Created a simulation that feels real - judges can visually see drones, boats, and trucks coordinating in real-time.
Brought inclusivity by adding multilingual communication, crucial in India where disaster victims may not understand English.
Demonstrated how AI can move beyond prediction to real-time autonomous action.

𝗪𝗵𝗮𝘁 𝘄𝗲 𝗹𝗲𝗮𝗿𝗻𝗲𝗱

How to leverage IBM’s ADK + Granite models for building agentic workflows.
The importance of human-AI collaboration — AI makes decisions, but humans need understandable, local-language outputs.
How simulation-first design can show a futuristic vision without needing full physical infrastructure.
Learned that coordination, not just prediction, is the key bottleneck in disaster management.

𝗪𝗵𝗮𝘁'𝘀 𝗻𝗲𝘅𝘁 𝗳𝗼𝗿 𝗥𝗲𝘀𝗤𝗡𝗲𝘁

Mobile App for Citizens → SOS reporting + receive real-time safety alerts in local languages.
Partnerships with NDMA & NGOs → Deploy at scale in disaster-prone regions (Assam floods, Odisha cyclones, Kerala).
IoT + Drone Integration → Incorporate live data from actual drones and ground sensors.
Global Expansion → Position this as a UN-level “𝗚𝗹𝗼𝗯𝗮𝗹 𝗔𝗜 𝗗𝗶𝘀𝗮𝘀𝘁𝗲𝗿 𝗦𝘄𝗮𝗿𝗺 𝗡𝗲𝘁𝘄𝗼𝗿𝗸”
AI Ethics + Trust Layer → Ensure explainability and safety of autonomous swarm decisions.