Abstract
This paper presents and analyzes a novel multiuser network based on hybrid free-space optical (FSO)/radio-frequency (RF) transmission system, where every user is serviced by a primary FSO link. When more than one FSO link fail, the central node services these corresponding users by using a common backup RF link according to a non-equal priority with $p$ persistence servicing protocol. A novel discrete-time Markov chain model is developed for the proposed network, where different transmission rates over both RF and FSO links are assumed. We investigate the throughput from central node to the user, the average size of the transmit buffer allocated for every user, the frame queuing delay in the transmit buffer, the efficiency of the queuing system, the frame loss probability, and the RF link utilization. The network's users are arranged in an ascending order. As the order of the failed user increases, the corresponding performance criteria gets worse due to the non-equal priority servicing protocol. Numerical examples show that transmitting a data frame with probability $p$ when using the common backup RF link, achieves considerable performance improvement that approaches the performance of the multiuser FSO/RF network when using equal priority protocol to serve all the remote users.
Introduction
Free-Space Optical (FSO) communication offers massive bandwidth and high security, but it is highly susceptible to atmospheric conditions such as fog, turbulence, and pointing errors. To maintain reliability, hybrid systems pair FSO links with more resilient, albeit slower, Radio Frequency (RF) links as backups. In multiuser scenarios, providing a dedicated backup RF channel for every user is often cost-prohibitive. Consequently, users must share a single RF link, necessitating an efficient scheduling protocol to manage access when multiple primary FSO links fail simultaneously.
System Model and Protocol
The authors propose a non-equal priority with $p$-persistence protocol to address the scheduling challenge. In this system, users are ranked by priority (1 to $N$). If multiple FSO links fail, the central node offers the backup RF link to the highest-priority failed user with a probability $p$. If that user is not served (due to the $1-p$ chance), the opportunity moves down the priority line.
To evaluate this, a sophisticated cross-layer Discrete-Time Markov Chain (DTMC) model was developed. This model accounts for physical layer dynamics (statistical fading for FSO and RF), data link layer queuing (finite transmit buffers), and rate asymmetry between the high-speed FSO and lower-speed RF links (defined by a factor $\Omega$).
Performance Analysis and Key Insights
The $p$-persistence parameter serves as a "fairness knob." A strict priority scheme ($p=1$) allows high-priority users to monopolize the backup link, potentially starving lower-priority users. By adjusting $p$, the system grants lower-priority users a statistical chance to access the backup, significantly improving their throughput and reducing latency.
Key insights from the analysis include:
- Throughput Dynamics: A characteristic "dip" in throughput occurs for lower-priority nodes as the frame arrival rate ($\omega$) increases, caused by buffer saturation while waiting for RF link access.
- Fairness vs. Priority: Implementing $p=0.5$ instead of strict priority dramatically improves the performance of lower-tier nodes, bringing them closer to an "equal priority" baseline without negating the benefits for high-priority users.
- Optimization: The model reveals that there is an "optimal $p$" that maximizes total network throughput depending on environmental conditions.
Conclusion and Future Work
The proposed DTMC model was validated through Monte-Carlo simulations, showing a near-perfect match. The study quantifies how RF link utilization saturates as the number of users increases, providing a tool for network dimensioning.
Future research could extend this work by relaxing the assumption of identical user fading statistics, exploring pools of multiple backup RF channels, or implementing dynamic priority levels based on real-time network state or data urgency.
Summary updated