A CLSTM and transfer learning based CFDAMA strategy in satellite communication networks

Area of Science:

• Satellite Communication
• Network Engineering
• Machine Learning

Background:

• Increasing demand for communication drives interest in satellite networks.
• Traditional satellite multiple access schemes do not account for data generated during delay times.

Purpose of the Study:

• To propose a novel multiple access scheme for satellite communication networks.
• To address data accumulation issues caused by communication delays.

Main Methods:

• Investigated a combined free/demand assignment multiple access (CFDAMA) scheme.
• Integrated convolutional long short term memory (CLSTM) network and transfer learning (TL) for data prediction.
• Developed a CLSTM-TL based prediction method (CLSTMTL) to forecast data generated during delay periods.

Main Results:

• The proposed CLSTMTL method predicts data generated during the satellite communication delay time.
• Incorporating CLSTMTL into CFDAMA reduces data accumulation by including predicted slots in requests.
• Simulation results demonstrate the effectiveness of CFDAMA-CLSTMTL compared to existing schemes.

Conclusions:

• The CFDAMA-CLSTMTL scheme offers an effective solution for managing data in satellite communication networks.
• The predictive capability of CLSTMTL enhances efficiency by mitigating data accumulation.
• This approach represents a significant advancement in satellite network multiple access schemes.