Abstract
Device-to-device (D2D) communication has seen as a major technology to overcome the imminent wireless capacity crunch and to enable new application services. In this paper, we propose a social-aware approach for optimizing D2D communication by exploiting two layers: the social network and the physical wireless layers. First we formulate the physical layer D2D network according to users' encounter histories. Subsequently, we propose an approach, based on the so-called Indian Buffet Process, so as to model the distribution of contents in users' online social networks. Given the social relations collected by the Evolved Node B (eNB), we jointly optimize the traffic offloading process in D2D communication. In addition, we give the Chernoff bound and approximated cumulative distribution function (CDF) of the offloaded traffic. In the simulation, we proved the effectiveness of the bound and CDF. The numerical results based on real traces show that the proposed approach offload the traffic of eNB's successfully.
Introduction
As mobile data consumption—driven by platforms like YouTube and Facebook—skyrockets, cellular networks face an imminent "capacity crunch." Often, multiple users in the same vicinity request the exact same popular content. In a traditional setup, the base station (eNB) sends duplicate transmissions to each user, wasting precious bandwidth.
Device-to-Device (D2D) communication allows users to share data directly over licensed bands, bypassing the eNB. However, D2D is tricky: physical connections are intermittent as people move, and it is difficult to predict which user has the content another user wants. This research addresses this by bridging the gap between the physical wireless layer (who is near whom?) and the social network layer (who wants what?).
Proposed Framework
The authors propose a two-layer model to optimize traffic offloading:
- Offline Social Network (OffSN): This models the physical layer. Instead of assuming users are always connected, the authors use encounter histories to estimate "closeness." They model the duration of user contacts using a Gamma distribution $\Gamma(k, \theta)$. If the probability of a contact lasting long enough to transfer a file exceeds a certain threshold ($w_T$), a D2D link is considered viable.
- Online Social Network (OnSN): This models content dissemination. The authors use the Indian Buffet Process (IBP)—a stochastic process typically used in non-parametric Bayesian statistics—to model how users select content. In this analogy, "dishes" are online contents and "customers" are users. This allows the eNB to predict the popularity of "old" content (already downloaded by someone in the vicinity) versus "new" content.
The eNB acts as a central coordinator, collecting encounter histories and social tags to decide whether to serve a request directly or facilitate a D2D transfer. To evaluate performance, the authors derive a Chernoff bound and an approximated Cumulative Distribution Function (CDF) to estimate how much traffic can realistically be offloaded.
Datasets and Methodology
The study validates its model using real-world traces from the CRAWDAD repository (University of St Andrews):
- Encounter Records: Data from 27 "T-mote" sensor devices deployed among users over 79 days to track physical proximity.
- Social Topology: Facebook friend lists from the same group of participants to map online social ties.
By clustering users into "OffSNs" based on social ties, the network can create more reliable D2D subnetworks. The IBP accounts for both "external influence" (trending videos) and "internal interest" (personal preference), providing a more realistic model for content requests than simple static popularity rankings.
Performance Evaluation
Simulation results based on real traces show that the proposed approach successfully reduces the eNB's traffic burden. Key findings include:
- Traffic Offloading: As the maximum allowable D2D communication distance increases, the eNB sum-rate decreases, indicating that more traffic is successfully offloaded to the D2D layer.
- Mathematical Accuracy: The derived Chernoff bounds and the Saddlepoint approximation for the number of "old" contents selected by users closely matched the empirical simulation results.
For a test case where $\alpha=20$, the mean number of offloadable contents was approximately 15, showing high efficiency in content reuse.
Conclusion and Future Work
This social-aware framework demonstrates significant potential for easing cellular network congestion. However, increasing D2D distance increases the "cost" for user equipment (UE), specifically in terms of battery power consumption. Future work could delve deeper into incentive mechanisms to encourage users to share content. Additionally, optimizing for inter-OffSN interference in extremely dense environments remains an open area for research.
Summary updated