Abstract
As cellular networks transition toward dense small cell deployments, the traditional benefits of opportunistic scheduling—selecting the user with the optimal channel condition—are diminished by traffic burstiness and low user density. This paper proposes a client-side "traffic spreading" mechanism where users cooperate via secondary radio interfaces to balance base station (BS) queues. By ensuring the BS always has multiple backlogged queues to choose from, the system restores opportunistic gains. Results demonstrate that this approach can reduce file transfer delays by up to 50% with minimal impact on battery life.
Introduction: The Challenge of Small Cells
Modern cellular base stations rely on opportunistic scheduling to maximize throughput by transmitting data to users with the strongest instantaneous signal. While effective in large macro-cells with many active users, this strategy falters in small cells (picocells or femtocells). In these environments, the small number of active users often leads to "empty queues" for those with the best channel conditions. When a high-quality channel goes unused because of a lack of data, the BS is forced to serve users with poorer signals, leading to a significant loss in overall network efficiency.
Proposed Method: BS-Transparent Traffic Spreading
To address this, the authors introduce a traffic spreading approach that operates entirely on the user equipment (UE), making it transparent to existing BS hardware. The core of the system is a "dispatcher" running on each mobile device. When a user initiates a download, the dispatcher evaluates whether to request the file directly or to leverage a nearby peer.
As shown in Figure 3, in the cooperative mode (b), User 2 ($U_2$) can send a request to User 1 ($U_1$). $U_1$ then forwards this request to the BS, which populates $Q_1$. The BS, seeing a backlog in $Q_1$, can then schedule transmissions to $U_1$ whenever $U_1$ has a superior channel. Once $U_1$ receives the data, it forwards it to $U_2$ via a short-range D2D link (e.g., WiFi). This process effectively balances the BS queues, providing the scheduler with more options to exploit favorable channel conditions.
Mathematical Modeling and Heuristics
The decision-making process for the dispatcher is modeled as a Markov Decision Process (MDP). The objective is to minimize a cost function that balances file transfer delay against the energy consumption required for D2D forwarding.
The system analyzes the request rates ($\lambda$) and queue states ($Q$) to determine the optimal direction for traffic spreading. Figure 6 illustrates how the dispatcher controls the "direction" of incoming requests to stabilize the system and maximize opportunistic gain. Because solving the MDP for many users is computationally intensive, the authors developed a scalable heuristic that decomposes the problem into pairwise interactions, allowing for real-time execution on mobile processors.
Key Results and Insights
Simulations based on LTE standards and Rayleigh fading channels reveal several critical insights:
- Significant Delay Reduction: Average file transfer delays were reduced by 20% in homogeneous scenarios and by up to 50% in heterogeneous environments where signal strengths varied significantly among users.
- Energy Efficiency: The system follows an "80/20 rule," where approximately 78% of the maximum possible gain is achieved using only 20% of the maximum energy budget. This makes the approach highly viable for battery-constrained smartphones.
- Scalability: The heuristic algorithm maintains performance within 2% of the optimal MDP solution while remaining light enough for deployment on standard mobile processors.
Conclusion
User cooperation through traffic spreading offers a powerful, low-cost solution to the inefficiencies of small cell scheduling. By shifting the logic to the client side, this method bypasses the need for expensive infrastructure upgrades. Future work will focus on incentive mechanisms to encourage altruistic behavior among users and optimizing performance in high-mobility scenarios.
Summary updated