Joint Transmission Rate and Cost Network Switching Mechanism Based on DQN for High-speed Vehicle

The high-speed vehicle causing high-frequency network switching (NS), surges signaling overhead for mobility management. As the scale of vehicle network continues to expand, NS cost from signaling overhead will effect the continuity of service transmission, which is not affected by transmission rate. To address this problem, this paper proposes a NS mechanism considering transmission rate and NS cost. First, it is based on DQN that can find target network online by using channel state information and location immediately. Second, this paper summarizes the key access points of vehicle network as Network Unit (NU), which doesn’t only cover the fundamental network architecture, but also is able to form a complete vehicle network. The mechanism operates in NUs scene. Simulation results show that, when a vehicle travels at 90 $\mathrm{km} / \mathrm{h}$ and service accesses randomly, average NS times reduces by $25 \%$, comparing with ignoring cost situation. In addition, the average NS times are reduced by $12.75 \%$ compared to the Q-learning algorithm with Convolutional Neural Network (CNN).These reflect the proposed mechanism reduces NS cost with transmission rate ensurance.