Satellite Beam Slicing Resource Allocation Strategy Based on Business Recognition and Prediction

Slicing resource allocation technology enables more flexible application of network resources from end to end. Traditionally, fixed air interface slicing configuration strategies such as resource reservation have been used to meet business demands that change slowly over time. Low Earth Orbit (LEO) satellite constellations, characterized by wide coverage, high capacity, and low latency, are being rapidly developed and integrated into space-terrestrial integrated networks. Due to the unique movement patterns of LEO satellites, business demands within different time periods and coverage areas vary dynamically, making fixed air interface slicing configurations inadequate for this dynamic LEO satellite scenario. This paper proposes an AI-based strategy for service identification, service prediction, and proactive slicing resource allocation. This strategy supports the advance adjustment of resource configurations according to changes in business demands, thereby enhancing the alignment between business demands and resource adaptability. The results show that the accuracy of service identification, based on the Pattern Recognition Toolbox, is no less than 98%. Additionally, the error of service prediction, leveraging Long Short-Term Memory (LSTM) neural networks, is no greater than 33%. Furthermore, the proactive slicing allocation optimizes the resource surplus rate and service insufficiency rate by 3.2% and 4.7%, respectively.