Neural Network-Aided Cell Selection and Handover in Cellular Networks with Carrier Aggregation

Carrier aggregation is one of the most important technologies for increasing data rates in current and future cellular network systems such as 5G and 6G. It combines multiple frequency channels, possibly from different bands such as bands under 6GHz and millimeter-wave bands, to send traffic between a base station and a user device at a faster speed. With carrier aggregation, a user device is simultaneously connected to multiple cells, one of them being the primary cell and the others being secondary cells. In environments where carrier aggregation is extensively used, the cell selection algorithm must take into account the performance of all potentially aggregated cells when selecting the best cell for a user device. However, the current cell selection method only considers the signal quality of primary cells, which may lead to sub-optimal performance when carrier aggregation is used. Modifying the cell selection algorithm to consider both primary and secondary cells may seem trivial, but faces new challenges such as increased measurement cost and delay. We present a simple and practical scheme for cell selection in carrier aggregation environments. First, the proposed scheme introduces new measurement configurations and events that can implement a cell group-based handover considering the combined performance of aggregated cells. Second, to reduce delay and overhead caused by measurements, we make use of neural networks that are trained to estimate the signal quality of neighbor cells at approximate user locations. Simulation results show that the proposed scheme achieves significant performance benefits over legacy methods, especially in dense network environments.