Idea
A deep learning platform using 5G GNSS signals to deliver accurate real-time 3D wind field retrieval and short-term forecasts for weather and aviation.
Research Paper
Core Innovation
This paper introduces G-WindCast, which uniquely leverages 5G GNSS signal strength variations combined with deep learning models to retrieve and predict 3D wind fields. Unlike traditional methods relying on in-situ sensors or computationally expensive numerical models, it captures complex nonlinear spatiotemporal wind dynamics efficiently. The approach also demonstrates robustness with fewer GNSS stations, enabling scalable and cost-effective localized forecasting.
Market Size (TAM)
$10–20B TAM for atmospheric and weather data services; $2–10B SAM from aviation, renewable energy, and disaster management sectors. Driven by increasing demand for real-time localized weather data and advances in 5G infrastructure.
Potential Customers & Pain Points
- Weather Forecasting Agencies Needing Higher Resolution Wind Data
- Aviation Safety Operators Requiring Real-Time Wind Information
- Disaster Risk Management Teams Lacking Rapid Localized Wind Forecasts
- Renewable Energy Firms Optimizing Wind Resource Assessment
- Telecommunications Providers Monitoring Atmospheric Conditions
Business Model
Subscription-based API and platform services for real-time wind data and forecasts targeting weather agencies, aviation, and energy companies.
Competitive Landscape
- The Weather Company
- Tomorrow.io
- ClimaCell
Implementation Challenges
- Integration with existing weather infrastructure
- Dependence on GNSS station density and coverage
- Validation across diverse geographic regions
Validation Strategy
- Pilot deployment with regional weather agencies
- Comparison against high-resolution NWP and ERA5 data
- Field testing with reduced GNSS station setups
Research Paper Overview
Communications to Circulations: 3D Wind Field Retrieval and Real-Time Prediction Using 5G GNSS Signals and Deep Learning
Summary
G-WindCast is a deep learning framework that uses 5G GNSS signal strength variations to retrieve and forecast 3D atmospheric wind fields with high accuracy and real-time capability. It combines Forward Neural Networks and Transformer models to capture complex spatiotemporal wind dynamics, achieving performance comparable to high-resolution numerical weather prediction models. The system remains robust across different forecast horizons and pressure levels and performs well even with fewer GNSS stations, making it scalable and cost-effective for localized wind forecasting.