Real-World Spatio-Temporal Behavior Aware D2D Multicast Networks

Studies show that real-world user mobility has certain spatio-temporal correlations. However, most of the existing works on device-to-device multicast (D2MD) networks use randomly generated user locations, thus fail to capture the real-world spatio-temporal behavior of the users. In this work, we propose a realistic performance evaluation of the D2MD network that takes into account the joint spatio-temporal correlations of the users in the network. The proposed performance evaluation will be beneficial in assessing the viability of the D2MD networks before their actual deployment. Further, in order to acquire the required joint spatio-temporal behavior, we propose two novel methods namely observed mobility exploitation (OME) and expected mobility exploitation (EME). In OME, the joint spatio-temporal behavior is extracted from the past location information of the users. On the other hand, in EME, the past location information of a set (i.e., the training set) of users, following an average spatial-temporal behavior similar to the intended users, is utilized to estimate the intended users' joint spatio-temporal behavior. The realistic performance of two D2MD networks is evaluated for a tractable campus set-up in terms of content delivery time and data offload ratio. It has been shown that for the D2MD networks the data offloading varies spatio-temporally and goes up to 35%. Further, a comparison is drawn between the conventional cellular unicast and multicast networks, and the two D2MD networks.