Research on mobility prediction in 5G and beyond for vertical industries

Abstract

Intelligent mobility management and service control based on UE mobility prediction is one hot topic in current AI-assisted 5G/B5G research, which can also be applied to vertical industries communication. We analyzed the standard progress in 3GPP and IMT-2020 on AI based mobility prediction, as well as the challenges and possible solutions. We selected three different models, which are Long Short-Term Memory (LSTM) model, Attention Bidirectional Long Short-Term Memory (BiLSTM-attention) model and Artificial Neural Network (ANN) model for AI based mobility prediction, using the data from Geolife project of Microsoft Research Asia as input. By model retraining and optimization, we achieved higher accuracies of trajectory prediction (around 90%) with shorter training time. In addition, we presented examples of applying UE mobility prediction in various verticals, for which our optimized models may be applied.