InfernoBots
Our project is an AI-powered autonomous emergency response system built entirely in simulation, demonstrating how intelligent multi-robot teams can detect, assess, and respond to wildfire and disaster scenarios with least human intervention. The system integrates an aerial drone (DJI Mavic 2 Pro) and ground robots operating inside a realistic Webots simulation environment. The drone performs real-time visual surveillance using a Vision-Language Model (VLM) to detect fire, smoke, damaged vegetation, and injured humans. Instead of following pre-scripted paths, the drone uses a orchestrated agentic flow to reason about the scene, decide where to move next, prioritize threats, and select appropriate actions such as dropping water bombs to suppress flames or deploying first-aid kits and breathing masks for victims or navigating to next target around the scene of criticality. The ground robot supports logistics and follow-up operations, navigating autonomously to transport essential equipment based on the aerial assessment. A centralized supervisory AI coordinates task allocation, ensuring efficient parallel execution while avoiding collisions and redundant coverage. The system emphasizes adaptive decision-making, dynamic mission planning, and perception-driven autonomy. All behaviors are designed to be transferable to real robotic platforms, showcasing how AI reasoning combined with robotic control can enhance disaster response operations in real-world environments.
Category tags:Additional links:
.png&w=96&q=75){kind=small}
"Description: AI-powered autonomous emergency response for wildfire/disaster scenarios. Aerial drone (DJI Mavic) + ground robots in Webots simulation. VLM detects fire, smoke, victims. Multi-robot coordination. Features: Aerial drone + ground robots VLM for fire/smoke/damage detection Water bomb dropping, first-aid deployment Autonomous navigation Centralized supervisory AI Demo: https://vercel.com/aasthaenggs-projects/inferno-bots Pros: ✅ EMERGENCY RESPONSE! - Different from warehouse robots ✅ Multi-robot coordination ✅ Real-world impact (wildfires) ✅ Working demo Cons: Zero stars"
Sanem Avcil
"Overview: This project demonstrates an AI-powered, fully autonomous multi-robot emergency response system simulated in Webots, designed for wildfire and disaster scenarios with minimal human intervention. A DJI Mavic 2 Pro-style aerial drone uses a Vision-Language Model (VLM) for real-time detection of fire, smoke, damaged vegetation, and injured humans. Pros: Include strong integration of modern VLMs and agentic reasoning for perception-driven autonomy, effective aerial-ground team coordination, high relevance to real-world disaster response, and a focus on simulation realism with transferability in mind. Cons: Potential weaknesses are the lack of physical hardware validation or Sim2Real experiments, simplified modeling of payload actions (e.g., drops), possible VLM robustness issues in smoke/low-visibility conditions, high computational demands for real-time edge deployment, and limited quantitative benchmarking against baselines"
Anton Kiselev
Lead Backend Developer
For Innovators & Creators