隨著衛星網路的逐漸崛起，越來越多研究試著將衛星網路融入現存的應用架構中。在這篇文章中，為了強化系統的覆蓋範圍並且在壅塞的車聯網路中保持良好的通道品質，我們試著整合衛星網路以及車連網路形成衛星協助多連結性車聯網架構。衛星協助多連結性車聯網架構乃指車輛能夠選擇要傳輸資料給衛星、基礎設施以及其他車輛。車輛能夠自行選擇使用何種傳輸模式、能量以及子頻道去最大化整體系統的效益。在我們的研究中，我們使用多智能體行為評斷模型 (MAAC) 去估計廣域狀態中的區域狀態。此外，我們使用城市交通模擬系統 (SUMO) 依照現實地圖去產生城市、郊區以及鄉村的車流量。根據研究結果顯示，我們方法能夠增加在鄉村的系統覆蓋範圍以及緩解都市的車用網路壓力。;With satellite network regaining the attention of the public, there are more and more research try to integrate the satellite network in nowadays application structure. In this paper, to enhance the system coverage in vehicle-to-everything (V2X), we expand the transmission target from infrastructures and other vehicles to satellite. The vehicle agent can arrange their transmission modes, power and sub-channel according to the environment to maximize the overall the system utility. However, the vehicle-to-satellite (V2S), vehicle-to-infrastructure (V2I) and vehicle (V2V) have different advantage and spectrum resources in different area. To maximize the utility of system in such complex environment, we apply the multi-agent-actor-critic attention (MAAC-A) which estimate the global state given partial information with attention mechanism to increase the learning efficiency. With MAAC-A, the vehicles can do the better selection with local information and maximize the utility of system immediately. Moreover, in our simulation, the data which including the urban, suburban and rural area is generated by Simulation of Urban Mobility (SUMO) on realistic setup. Finally, the result show that the agent has advantageous performance with the proposed scheme without the satellite network.
