Flow control input prediction for satellite based internet service

Abstract

In a satellite communication system, a method and system for proactive bandwidth allocation are introduced. The method involves collecting data from input producers and a flow controller, tracking real-time traffic inputs, and predicting future network traffic patterns using a Flow Control Input Farseer (FCIF) model. This model is trained with historical traffic data and utilizes neural networks such as RNNs, LSTMs, and CNNs. The system includes a Farseer Agent (FA) for data collection, input producers for tracking traffic, and a flow controller for bandwidth assignment based on predicted traffic patterns. The FCIF model is refined by a trainer and incorporates feature engineering and attention mechanisms to enhance prediction accuracy. The system also includes an input aggregator for managing traffic reports and bandwidth requests.

Dependencies are inferred from the claim text.