Introduction
In modern Time Division Duplex (TDD) networks, such as 5G, the same frequency band is shared for both uplink and downlink transmissions. This shared medium introduces significant challenges for network modeling, particularly when the Base Station (BS) serves as a relay for User Equipment to User Equipment (UE2UE) communication. This paper introduces a novel framework for determining the "stability region"—the maximum arrival rates a network can sustain without queues growing indefinitely. By transforming complex, multi-dimensional queue interactions into simplified one-dimensional models, the authors provide a method to calculate network capacity that balances mathematical precision with computational efficiency.
The Challenge of Queue Coupling
A primary difficulty in modeling TDD networks is the phenomenon of queue coupling. In a relay scenario, a packet cannot be transmitted from the BS to a destination user until it has been successfully received by the BS from the source user. Consequently, the performance of the downlink queue is intrinsically linked to the state of the uplink queue.
Traditional research often simplifies this by assuming saturated queues or using basic ON/OFF models. However, real-world traffic is bursty and radio conditions are dynamic. Modeling these conditions exactly typically leads to a "state space explosion," making the analysis computationally prohibitive for large-scale networks.
Methodology
To overcome the complexity of standard multidimensional Markov Chains, the authors employ several innovative techniques:
- Priority Policies: By applying priority-based scheduling, the multidimensional problem is decomposed into a set of one-dimensional (1D) Markov Chains. This allows the stability region to be characterized as a convex polytope defined by a finite number of vertices.
- Approximated Markov Chain: A simplified "birth-and-death" Markov Chain is proposed for the Base Station queue to avoid the computational burden of solving massive systems of linear equations.
- Epsilon-Close Approximation: The authors derive an analytic upper bound for the stability region. This "epsilon-close" model offers a closed-form expression that significantly accelerates computation while maintaining a strictly bounded error margin ($\epsilon$).
- Complexity Reduction: A parameter ($K_0$) is introduced to limit the number of scheduling policies required to achieve a target level of precision in multi-user scenarios.
Performance Evaluation and Key Insights
The framework was validated through numerical simulations of a 10 MHz LTE-like TDD system. Key findings include:
- Queuing Impact: Analysis reveals that physical layer metrics alone are insufficient for performance evaluation. Incorporating queuing theory shows that an empty relay buffer can actually improve uplink service rates, as the scheduler does not allocate resources to an empty downlink.
- Computational Efficiency: The most significant result is the reduction of computational complexity from exponential to polynomial. For a 50-user network, the required operations are reduced from approximately $3 \times 10^{79}$ to $9 \times 10^5$.
- Precision and Scalability: The proposed approximation serves as a tight upper bound, with numerical results showing nearly identical curves for exact and approximated stability regions across various distances (100m to 500m).
Conclusion
This research provides a scalable and precise framework for analyzing the stability of TDD networks. By effectively managing the trade-off between complexity and precision, the model allows for practical network capacity planning. Future work will focus on extending these models to non-symmetric deployments, OFDMA systems, and integrated external downlink traffic.
Summary updated