An LSTM-based cell association scheme for proactive bandwidth management in 5G fog radio access networks

Cellular networks are evolving into dynamic, dense, and heterogeneous networks to meet the unprecedented traffic demands which introduced new challenges for cell resource management. To address this, various cell association schemes have been proposed. However, the current schemes ignore users' mobility information, and as a result, their cell admission and bandwidth allocation policy are reactive. In order to enable proactive bandwidth management in emerging 5G fog access radio networks (FRANs), we proposed a novel mobility-aware cell association scheme (MACA) that exploits user's mobility and downlink rate demand information to associate it with the maximum rate offering cell. In MACA, the mobility prediction model consists of long-short-term memory (LSTM)-based neural network that considers joint information of unique cell identification numbers with the corresponding sojourn times and predicts the user's most probable next cell. Later, the underlying future cell assignment is formulated as a convex problem and solved using the Lagrangian dual decomposition method and compared the proposed framework performance with Semi-Markov, deep neural network (DNN), and MaxRSRP-based cell association approaches in terms of the next cell prediction accuracy, the impact of downlink rate allocation, and user satisfaction percentage. MACA scheme is trained using two publicly available pedestrian datasets. Simulation results show that the proposed scheme performs significantly better than the other schemes and yields the average next cell prediction accuracy of 93.42%, 1.63 times higher downlink rates, and 56.8% users satisfied with the allocated bandwidth.