| Summary: | 碩士 === 國立中正大學 === 通訊工程研究所 === 105 === As the number of users of LTE-A grows rapidly, the operator faces the challenge of wireless resource scarcity. The device to device (D2D) communication has been regarded as a valid way to mitigate this resource scarcity problem. In the meantime, with the booming of video streaming services, the eNB needs to use multicast for higher resource efficiency. When the eNB performs multicast to all users under its serving cell, it needs to consider the worst channel gain user for the multicast to be successful. In this situation, the eNB is forced to use a poor modulation scheme to serve all users.
In this thesis, we consider using D2D communication to assist the eNB doing multicast. Since D2D communication works in a short range, the original eNB multicast is broken into two shorter-range multicast segments and a better modulation scheme can be used for both segments, resulting in higher multicast throughput. On the other hand, in assisting multicast, a short-range D2D communication cannot serve all users in the same cell. Therefore we need an effective clustering scheme to group users into multiple clusters, each with geographically close users.
In addition, when we want to use D2D communication, we need to choose a wireless channel appropriately. In the LTE-A standard, in contrast to the uplink resources, many downlink resources are allocated for delivering the control signals. For this reason, we consider using the uplink resource for D2D to share with cellular users. In earlier studies, a resource unit is shared between one cellular user and one D2D communication user. But in this thesis, we let a cellular user share the uplink resource with multiple D2D communication users. The feasible resource sharing is found by the optimization model and suitable clustering.
From the simulation results, we find that, when the cellular user shares the uplink resource with multiple D2D communication users, it is possible to save more resources without the loss of throughput for D2D communication. Also, we compare the difference among three clustering approach: Fuzzy C-means, K-means++, and Affinity Propagation. According to the simulation result, we find that the Affinity Propagation is more suitable for the resource allocation scheme we propose.
|
|---|
