Page 388 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

Use Case 20: AI-Driven Free Space Optical Communication System

Country: Tanzania

Organization: Mbeya University of Science and Technology (MUST)

Contact Person(s):

Silvester Beatus Mhagama, Email: silv estersilve ster36362@ gmail .com
Dr. Joseph Sospeter, Email: joseph.sospeter@ must .ac .tz

1 Use Case Summary Table

Item Details

Category 5G
Problem Addressed Severe signal degradation in Free-Space Optical (FSO) communication
due to weather conditions (fog, rain, turbulence).
Key Aspects of Solu- AI-driven dynamic power control, hybrid FSO/RF switching, real-time
tion weather data integration, adaptive modulation

Technology Keywords AI/ML, Optical Communication,
Atmospheric Modeling, IoT Integration, 5G/6G

Data Availability Public: TMA weather datasets.
Private: Simulation data from MUST labs.

Metadata (Type of Time-series visibility, rain rate, humidity; signal strength (dBm), BER logs.
Data)

Model Training and LSTM networks for weather prediction; reinforcement learning for adap-
Fine-Tuning tive power optimization.
The AI component uses time series models to predict optical signal
attenuation over time. We are currently experimenting with LSTM (Long
Short-Term Memory) and Prophet models to forecast the signal strength
based on environmental variables

Testbeds or Pilot MATLAB Simulink prototype.
Deployments We verify the AI predictions against simulated optical signal data in
MATLAB Simulink, using the actual weather conditions as input. Metrics
such as RMSE (Root Mean Square Error) and BER (Bit Error Rate) were
used to validate accuracy.

Code repositories AI-driven FSO/mwanasheria/GitHub available on GitHub

352