近年來智慧型行動裝置的普及帶動了行動資料流量快速成長，為因應行動數據流量呈現爆炸性的增長，行動電信營運商試圖將網路核心設備升級來擴充服務數據容量，但相較於呈指數倍增的行動數據流量而言，其成效相當有限。此外，目前多重網路整合之相關研究，多數需用戶端與伺服器端皆支援所提之協定，或有封包重置(packet reordering)形成傳輸效率瓶頸之問題。是以本論文提出了基於SDN之多重無線網路頻寬聚合機制， 來處理LTE與Wi-Fi等多重網路間的傳輸路徑選擇與切換。所提之多重連線接取軟體、聚合控制應用、與資料流分析應用分別安裝於UE端與SDN網路端以處理多重無線網路頻寬聚合。藉由我們提出的機制，UE端僅需安裝多重連線接取軟體，而原本執行於UE端的應用服務不需做任何修改即可使用。則當單一網路無法滿足UE之QoS需求時(如UE同時執行多個需要不同等級服務品質之應用服務)，可經由本論文提出之SDN多重無線網路頻寬聚合機制， 依據當前網路傳輸效能與UE欲傳輸之服務流特性進行傳輸路徑配置，滿足網路中各UE之網路服務品質需求，並提高整體網路之使用效率。

Nowadays, the popularity of smart mobile devices drives the mobile traffic growth rapidly. For handling the extreme growth of mobile traffic, mobile telecommunications operator tries to upgrade the core network equipment to expand the network capacity of services, but compare to the exponentially growing mobile traffic, the effect is limited. Moreover, most of the current multi-RAT integration researches need both the clients and servers support their proposed protocols, or have the transmission efficiency problem caused by the packet reordering. Hence this paper proposes the SDN-based multi-RAT bandwidth aggregation mechanism, for handling the transmission path selection and switching between LTE, Wi -Fi, and other multi-RATs. The proposed multi-RAT Adapter software, aggregation control application, and flow analysis application are installed in UE and SDN side respectively for handling the multi-RAT bandwidth aggregation. Through our mechanism, the UE only needs to install the multi-RAT Adapter software, and the application services executed in the UEs do not need to be modified under the proposed approach. When single network could not satisfy the QoS requirements of UE (e.g., the UE executes several application services which require different QoS classes simultaneously), by our SDN-based multi-RAT bandwidth aggregation mechanism, we can allocate the transmission path according to the current network transmission efficiency and the characteristics of the service flows transmitted by the UE. Thus we can satisfy the network QoS requirements of UEs and raise the efficiency of whole network utilization.