AWS Builder Center

Builder Center

AquaTread Bharat 🌊 — How We Built Vehicle-Aware Navigation with AWS AI

AquaTread Bharat 🌊 — How We Built Vehicle-Aware Navigation with AWS AI

AquaTread Bharat is a vehicle-adaptive emergency navigation PWA that reroutes drivers around flood hazards based on engine exhaust clearance and overall traffic, verified by Amazon Rekognition AI and live rain radar checks.

AquaTread Bharat 🌊 — How We Built Vehicle-Aware Navigation with AWS AI

Author: Devansh Upadhyay Team: ThunderStrike (Devansh Upadhyay, Hiya Khanna, Ayati, Srishti) Institution: JSSUN, Noida Hackathon Track: AWS Bharat Builds Hackathon 2026 — Environmental & Climate Track


💡 1. The Real Problem We Saw On Delhi-NCR Roads

Every monsoon in Noida, Delhi, and Gurgaon, my teammates and I see the exact same disaster unfold. A heavy 4x4 SUV or a bus wades through a flooded underpass with 40cm of water. Traditional navigation apps like Google Maps or Waze only measure vehicle movement speed. Because the SUV is moving, the app marks the road as "moving traffic" or "clear path".
A Sedan or a Scooter follows the app's guidance, enters the submerged road, and douses its engine exhaust. The vehicle hydro-locks, the driver gets stranded, and the entire stretch turns into a massive multi-hour gridlock.
Traditional navigation tools fail during Indian monsoons because they are completely blind to ground-level water depth and vehicle ground clearance.


⚡ 2. What We Built: AquaTread Bharat

We built AquaTread Bharat as an instant, zero-dependency Progressive Web Application (PWA). Instead of routing everyone purely on speed, AquaTread calculates engine-safe routes based on your ride's ground clearance.

Core Features:

  1. Vehicle Exhaust Clearance Routing Engine:
  • 🛵 Two-Wheeler (15cm limit): Automatically avoids ankle-deep water logging.
  • 🚗 Sedan (30cm limit): Automatically avoids knee-deep water logging.
  • 🚙 SUV / 4x4 (45cm limit): Navigates deeper water safely.
  1. Dynamic OSRM Detours: When a flood hazard is reported along your path, the app automatically finds candidate waypoints and calculates a safe and fast possible detour bypassing the water logged zone.
  1. Heatwave Safe-Haven Mode: During extreme summer months, toggling Heatwave mode queries the Overpass API to map real-time OpenStreetMap drinking water points within a 3km radius.
  1. Transparent Live Traffic Overlay: Google traffic speed vectors sit cleanly inside our bright aqua route border on a dedicated Leaflet pane ( z-index: 465 ), keeping dark mode map colors crisp without clutter.


☁️ 3. Our AWS Cloud Architecture

We needed reliable image verification for crowdsourced flood reports without running heavy servers 24/7 or risking unexpected cloud costs.
  • AWS Lambda ( backend/aws_rekognition_handler.py ): A serverless Python 3.11 function that receives base64 image payloads with a strict 5MB payload safeguard.
  • Amazon Rekognition: Calls  detect_labels  with  MinConfidence=70.0  to detect tags like water ,  flood ,  puddle , and  submerged .
  • Amazon API Gateway: Exposes a CORS-enabled REST endpoint with proper preflight  OPTIONS  handling.
  • Client-Side Canvas Pre-Scanner: Runs RGB variance checks on the user's browser before triggering any cloud API.
  • Open-Meteo Weather Radar: Fetches real-time precipitation ( mm/hour ) at reported GPS coordinates to cross-verify live storm activity.


🚀 4. Why AWS Cloud Was a Total Game-Changer For Us

To be completely honest, as a 4-person student team building this during a fast-paced hackathon, AWS was an absolute lifesaver. Here is why AWS made this project possible:
  1. Zero Model Training with Amazon Rekognition: Without Amazon Rekognition, we would have spent weeks collecting thousands of flood images, annotating data, renting expensive GPU instances (like EC2  g4dn ), and fighting custom PyTorch model drift. Rekognition gave us production-ready, highly accurate computer vision out of the box through a simple API call.
  1. Perfect Monsoon Elasticity via AWS Lambda: Climate emergency apps face unpredictable traffic spikes — zero load on sunny days, but hundreds of concurrent scans during a sudden cloudburst. AWS Lambda's serverless architecture handled this effortlessly, scaling from 0 to hundreds of executions in milliseconds, giving us 100% uptime while costing us exact ZERO on idle sunny days.
  1. Incredible Developer Experience with  boto3 : Interfacing with AWS Cloud AI took just 4 lines of Python code using  boto3 . The developer experience was so seamless that we didn't have to waste time managing servers or configuring environments — we could focus entirely on solving real-world flood routing logic.
AWS gave us enterprise-grade AI and serverless reliability in hours rather than months!


🥊 5. What Fought Back (Engineering Challenges & Solutions)

Building a real-time climate emergency app wasn't smooth sailing. Here are three main engineering hurdles we ran into and how we solved them:

Challenge A: Protecting AWS Quotas from Spam Uploads

Users could upload indoor selfies, dry road photos, or blue clothing to spam the system. If every upload went straight to the cloud, our AWS Rekognition quota and budget would drain instantly.
  • How We Fought Back: We built a client-side HTML5 Canvas CV Scanner. Before sending any data to AWS, the user's phone browser inspects image pixels for skin tone ratios (>12%), blue fabric saturation ( b - r > 20 && b - g > 12 ), and dry asphalt neutral grey tones ( Math.abs(r-g)<=15 && Math.abs(g-b)<=15 ). Unrelated photos get rejected locally at zero AWS cost.

Challenge B: Waterlogging When It Isn't Raining (Drain Overflow / Pipe Bursts)

During sewage line collapses or post-storm standing water, Open-Meteo weather radar returns  0.0 mm/hr  rain, which initially caused our system to reject legitimate reports.
  • How We Fought Back: We updated our verification logic. If the weather radar shows 0mm rain but both Canvas CV and Amazon Rekognition confirm waterlogging, the app dynamically tags the hazard as  Standing Water / Drainage Failure  instead of rejecting it.

Challenge C: Map Overlay Z-Index Clutter

Adding Google Traffic tiles inverted our dark mode map colors and obscured the primary blue navigation polyline.
  • How We Fought Back: We engineered custom Leaflet map panes:  .leaflet-overlay-pane  ( z-index: 460 , blue route weight 8) and  .trafficVectorPane  ( z-index: 465 , traffic line weight 5.0). Traffic lines now sit cleanly inside the glowing blue route border.


📊 6. How AquaTread Compares Against Legacy Apps

When we were testing AquaTread against existing tools during the hackathon, we realized how much was missing in standard map apps during monsoons. Here is a quick breakdown of what legacy apps do versus what we built:
Capability
Google Maps / Waze
MapMyIndia
AquaTread Bharat ⚡
Vehicle Exhaust Clearance Check
❌ No
❌ No
✅ YES (15cm / 30cm / 45cm)
Multi-Layer AI Photo Verification
❌ No
❌ No
✅ YES (Canvas + AWS Rekognition)
Live Precipitation Radar Check
❌ No
❌ No
✅ YES (Open-Meteo API mm/hr)
Heatwave Safe-Haven Water Points
❌ No
❌ No
✅ YES (Overpass API)
Zero-Load Offline-Ready PWA
❌ App Download
❌ App Download
✅ YES (Single File Instant Load)


🎯 7. Summary & Future Impact

AquaTread Bharat bridges the gap between ground-level water depth and dynamic vehicle-aware navigation. Powered by Amazon Rekognition AI and AWS Lambda, it protects drivers from monsoon drownings while scaling effortlessly across India's urban road networks during severe weather events.


👥 8. Built By Team ThunderStrike

Developed with passion by undergraduate engineering students at JSSUN Noida:
  • Devansh Upadhyay — Full-Stack PWA Navigation Engine & Dynamic Leaflet/OSRM Integration
  • Hiya Khanna — Client-Side Canvas CV Pixel Scanner & Anti-Fake Filters
  • Ayati — Open-Meteo Weather Radar & Overpass API Water Points Integration
  • Srishti — AWS Lambda Backend ( aws_rekognition_handler.py ) & Cloud Architecture
Try our live app at  https://devanssh01.github.io/AquaTread-Bharat-builds-/  or check out our code on GitHub at  https://github.com/devanssh01/AquaTread-Bharat-builds- .
# aws-lambda# amazon-rekognition# serverless# artificial-intelligence# web-development
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article