Performance Evaluation of Distance-Based Handover Using LSTM in LEO-Based NTN

Abstract

In terrestrial networks (TN), the user equipment (UE) identifies the cell edge by detecting differences in reference signal received power (RSRP). However, in non-terestrial networks (NTN), these effects are minimal due to the high altitude, and frequent handover (HO) may occur. For address this issue, the 3rd Generation Partnership Project (3GPP) Release-17 introduced novel HO triggering conditions, such as distance-based and time(r)-based HO [1]. Especially, distance-based HO addresses the challenge of minimal RSRP differences in cell overlapping areas. However, it is performed based on distance without considering RSRP, which can present challenges in maintaining the average RSRP of the UE. Therefore, a method is necessary to determine the optimal distance-based offset that can maintain RSRP for the UE. In this paper, we propose a distance-based HO using long short-term memory (LSTM) to predict the average signal-to-interference-plus-noise-ratio (SINR), identify the optimal HO triggering position, and determine the distance-based HO triggering offset. Performance analysis demonstrates that the proposed distance-based HO maintains the average SINR while significantly reducing unnecessary HO compared to the RSRP-based HO.