An Efficient Predictive Handover in LEO Satellite Networks using LSTM

Modern telecommunication systems rely on satellite networks to provide global coverage, especially in remote areas. Handover in Low Earth Orbit (LEO) satellite networks raises significant new challenges. Conventional handover algorithms rely on fixed, predefined rules and static thresholds to make a handover decision. However, these static handover algorithms often become inefficient under the dynamics of satellite movements and time-varying network conditions. Hence, we argue that it would be beneficial to measure the proximity of satellites to the specific cellular base station. This relative proximity helps in optimizing the handovers proactively. This results in the improved service quality and lower latency. We formulate a viable way of enhancing satellite handovers using distance-based predictive models and incorporating satellite bandwidth and path loss. Simulation results on actual satellite data shows more than $40 \%$ improvements in handover latency while ensuring the user receives adequate signal strength to maintain the network connectivity and minimizing the handover failure by $80 \%$.