A deployable AI-powered autonomous drone that aids search-and-rescue operations by detecting people and hazards, thereby improving responder safety and reducing victim discovery time.
“Disasters like floods and earthquakes make areas unreachable for human rescue teams, delaying crucial help. We need to build an autonomous drone that flies into these zones without human control, uses onboard cameras and AI to spot survivors and hazards, and maps everything offline. This gives emergency responders instant life-saving intelligence without risking their own lives.”
Deconstructed Problem Blueprint
WHAT TO BUILD?
An edge-AI powered autonomous drone hardware and software ecosystem comprising an onboard compute module (like Qualcomm RB5/RB3 or Jetson), dual RGB-Thermal camera setup, SLAM navigation system, and a command center ground control dashboard.
WHO IS IT FOR?
Primary beneficiaries are disaster victims trapped in remote or dangerous terrains. Primary users are National Disaster Response Force (NDRF), state disaster management authorities, and first responders.
WHY DOES IT EXIST?
During the first golden hours of disasters, communication lines collapse, maps become inaccurate due to debris changes, and manual ground reconnaissance is dangerously slow and places rescue workers at extreme risk.
INPUT → PROCESS → OUTPUT ARCHITECTURE PIPELINE
RGB camera frames, thermal video streams, IMU sensor telemetry, GPS coordinates (and optical flow/LiDAR for GPS-denied zones), and pre-loaded offline regional vector maps.
Onboard sensor fusion, real-time object detection and hazard classification using quantized neural networks, visual SLAM for simultaneous mapping and localization, and local geospatial indexing.
Live geo-tagged hazard and survivor markers, automated high-priority distress alerts, safest-route recommendations, and a synchronized situational report dashboard on the ground station.
CONSTRAINTS
- Must operate completely offline without cloud connectivity or standard GPS in deep canyons or collapsed structures.
- Strict power and weight budget to maximize drone flight time (endurance).
- Real-time processing constraint: Onboard AI inference must run at high FPS on edge silicon without overheating.
- High accuracy required under adverse weather (rain, dust, smoke, complete darkness via thermal spectrum).
JURY METRICS
- Onboard AI inference latency (Frames Per Second / millisecond delay per frame).
- Detection accuracy (mAP for humans, fire, smoke, and debris in thermal and RGB).
- SLAM trajectory drift and localization accuracy in GPS-denied environments.
- Dashboard UX response time and clarity of prioritized alerts during stress tests.
CORE BOTTLENECK
Achieving real-time, low-power multi-model AI inference (combining object detection + thermal anomaly detection + SLAM navigation) simultaneously on a resource-constrained edge flight controller without battery drain or thermal throttling.
Why You Should Choose This PS (Pros)
- Extremely high social impact addressing a real, urgent national security and disaster management priority.
- Directly backed by Qualcomm, opening opportunities for using their edge-AI hardware kits (like Qualcomm robotics platforms).
- Clear hardware-software integration pipeline with an impressive 'wow factor' during live demonstrations.
Challenges & Risks to Watch (Cons)
- Building physical drone hardware under hackathon constraints is risky and prone to crashes or hardware failures.
- Sourcing reliable thermal cameras and simulation environments (like Gazebo/ROS) requires prior robotics experience.
- Crowded domain if teams rely purely on standard YOLO wrappers without deep edge-optimization.
Recommended Tech Stack & Grand Finale Features
Killer Features to Impress the SIH Jury
View Official SIH 2026 Portal Raw Text▼
Background India is highly vulnerable to natural disasters including floods, cyclones,earthquakes, landslides, and flash floods, which often result in damaged infrastructure, inaccessible terrain, and delayed rescue operations. During the first few critical hours after a disaster, responders need rapid situational awareness to locate survivors, assess hazards, and prioritize rescue efforts.Traditional ground-based assessments can be slow, dangerous, and resource-intensive, particularly in remote or heavily damaged areas. Autonomous drones equipped with on-device AI can provide real-time aerial intelligence while operating in environments with limited connectivity.
Description Develop an autonomous drone system capable of navigating disaster-affected areas and performing real-time detection of survivors and hazards using on-device AI. The drone should use RGB and thermal cameras to identify stranded individuals, detect signs of human presence, and recognize environmental hazards such as fire, floodwaters, damaged structures, exposed electrical lines,debris, landslides, or chemical leaks. The solution must process data locally on the drone to ensure low latency and continued operation even when network connectivity is unavailable. The drone should autonomously map affected regions,generate situational reports, and transmit actionable insights to emergency response teams. This concept aligns with existing edge-AI drone approaches for incident response and disaster assessment.
Expected Solution The proposed solution should include some or all of the following: