Page 370 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Partners

International Telecommunication Union (ITU): Provides the foundational propagation models
(Recommendations P.530 and P.840) and guidelines for network resilience.

2�2 Benefits of the use case

The proposed system leverages real-time meteorological data, RNN-based forecasting, and RL
to proactively mitigate signal degradation caused by weather events. This innovation enhances
the resilience and adaptability of telecommunication infrastructure, especially in climate-
vulnerable regions. By introducing dynamic and intelligent control mechanisms, the system
promotes technological advancement, digital transformation, and robust industry growth.

By maintaining reliable network connectivity during adverse weather conditions, the system
safeguards access to critical services such as emergency response systems, telemedicine,
online education, and smart transportation. This ensures that both urban and rural populations
remain connected and resilient, even during extreme weather events. The ability to maintain
uninterrupted communication contributes to urban sustainability and community preparedness.

Weather-induced signal attenuation is a direct impact of climate variability, particularly in the
use of high-frequency mmWave bands. By integrating AI models that adapt to and mitigate
these effects, the solution acts as a form of climate adaptation technology. It supports the
continuity of critical infrastructure under climate stress, helping to reduce system vulnerability.
The proposed system is inherently collaborative, relying on data and coordination between
telecom providers, meteorological services, AI researchers, and public agencies. This cross-
sectoral integration exemplifies the call for multi-stakeholder partnerships to leverage
technology for sustainable development. By aligning telecommunications resilience with
environmental data and national digital goals, the project fosters global cooperation, capacity
building, and knowledge sharing.

2�3 Future Work

Model Development: Building on the initial concept guided by ITU’s P.530 [3] and P.840 [2],
we will enhance system accuracy by fusing physics-based propagation models with advanced
AI techniques. Forecasting accuracy for weather-induced signal degradation will be improved
using multi-modal inputs (3D Doppler radar, NOAA weather feeds, IoT sensor data). Metadata
extracted from these sources (timestamp, GPS coordinates, elevation, precipitation rate,
humidity, temperature, wind speed, and per-beam RSSI) will be formatted in CSV for batch
training and JSON for real-time streaming.

As preparatory work, we will evaluate the DeepSense 6G Beam Prediction toolkit [4], analyzing
its multi-modal data fusion techniques, benchmarking predictive performance, and extracting
best practices for feature engineering to inform our model pre-training strategy and metadata
schema design.

Reference Tools and Simulation Environment: An open-source toolkit and simulation platform
will be developed for reproducible testing and validation. We will use ns-3 with the mmWave
module and WeatherLab toolkit to generate synthetic impairments (rain rates 0–25 mm/h)
on virtual cell sites with 64-element arrays [5]. The AI inference pipeline will leverage Python

334