DQN-Based Vertical Handover Strategy for Satellite and Terrestrial Integrated Vehicular Networks

The terrestrial network (TN) can provide high data transmission rates and low-delay network services for vehicle terminals. However, base stations in the terrestrial network are typically stationary and have limited coverage. When vehicles begin to move at high speeds or travel to remote areas without coverage, frequent handovers may occur, reducing network service quality, or even causing service interruptions. In the satellite-terrestrial integrated heterogeneous network, supplementing the terrestrial network with Low Earth Orbit (LEO) satellite constellations can mitigate this issue. However, problems such as high speed, high delay, and wide coverage of LEO satellites can degrade the quality of vehicle service. Therefore, this paper constructs experimental scenarios based on real geographical environments. After fully considering the end-to-end (E2E) delay of the ground network and the impact of non-ground networks on handover decisions, a deep reinforcement learning-based handover strategy is proposed. A multi-attribute decision-making method is adopted to select the base station providing services according to different service preferences. Simulation results show that this method can effectively reduce the system's handover rate, ping-pong rate, and short Time-of-Stay (ToS) under different service distributions.