Algoritmo Híbrido de Handover com LSTM para Redes Open RAN com mobilidade Urbana

The densification of mobile networks and the Open RAN architecture impose challenges on mobility management, especially due to the increase in ping-pong handovers and unnecessary handovers that degrade the user experience. This paper proposes a hybrid handover algorithm that combines conventional methods and artificial intelligence to mitigate these issues. The approach uses the A2A4 event for the initial trigger and a Long Short-Term Memory (LSTM) neural network, in both multicell and single-cell versions, to predictively validate whether the target cell will maintain superior quality compared to the serving cell over the next 10 time steps. The solution was evaluated through integrated ns-3/SUMO simulations using the map of Juiz de Fora to represent a realistic urban scenario. The results show a reduction in ping-pong handovers and unnecessary handovers, highlighting the potential of the proposed solution for use as an xApp in Open RAN environments.