Abstract
This paper presents a rigorous mathematical framework for calculating the Extended Delivery Time (EDT) in cognitive radio networks characterized by non-work-preserving packet transmissions. By deriving exact probability distributions for various sensing scenarios—including continuous, perfect periodic, and imperfect periodic sensing—the study enables precise prediction and optimization of queuing delays for secondary users. Analytical results are validated via Monte Carlo simulations, providing a robust tool for designing latency-sensitive opportunistic spectrum access systems.
Introduction
In cognitive radio (CR) environments, secondary users (SUs) opportunistically access spectrum gaps left by primary users (PUs). In an interweave paradigm, SUs must cease transmission immediately upon PU reappearance to prevent interference. A critical challenge arises from the non-work-preserving nature of many communication protocols: if a packet transmission is interrupted, the partial data is discarded, and the SU must retransmit the entire packet from the start once the channel becomes idle again.
This interruption cycle leads to an "Extended Delivery Time" (EDT), which encompasses the actual transmission time, time wasted on failed attempts, and waiting periods during PU activity. While prior literature often focuses on mean delays, this work derives the exact statistical distributions (PDF and CDF) of EDT, which is vital for applications with stringent Quality of Service (QoS) requirements, such as industrial automation or healthcare monitoring.
System Model and Methodology
The authors model PU activity as a Markovian process with exponentially distributed busy and idle periods. The SU's interaction with the spectrum is analyzed under three sensing strategies:
- Continuous Sensing: The SU monitors the channel constantly, initiating transmission as soon as it becomes idle.
- Perfect Periodic Sensing: The SU checks the channel at fixed intervals ($T_s$).
- Imperfect Periodic Sensing: Accounts for sensing errors, specifically missed detections where a free channel is incorrectly identified as busy.
The mathematical derivation utilizes Moment Generating Functions (MGFs) for the constituent delay components. These MGFs are inverted using generalized hypergeometric functions to obtain the Probability Density Functions (PDFs). The resulting EDT distributions are then integrated into a generalized M/G/1 queuing model to evaluate the impact of continuous packet streams.
Key Insights into Extended Delivery Time
The EDT is conceptualized as a sequence of random "wasted slots" and "waiting slots" culminating in a single successful transmission slot.
- Wasted Slots: Occur when a PU interrupts an ongoing SU transmission.
- Waiting Slots: Represent the time an SU must wait for the PU to vacate the channel or for the next sensing interval to occur.
The choice of sensing interval ($T_s$) presents a fundamental trade-off: longer intervals reduce energy consumption but increase the "dead time" between the PU's departure and the SU's detection of the idle state. Furthermore, the performance gap between work-preserving and non-work-preserving strategies is shown to narrow as packet sizes decrease, though non-work-preserving systems remain significantly more sensitive to PU activity.
Results and Discussion
Validation through Monte Carlo simulations confirms a perfect match between the derived analytical formulas and simulated data across various parameters, including packet size and sensing error rates.
- Sensing Errors: Even minor probabilities of missed detection ($p_e$) significantly shift the CDF of the EDT to the right, indicating substantial increases in delivery time.
- Queuing Dynamics: The application of EDT to an M/G/1 queue reveals that PU activity effectively "stretches" the service time for SUs. Non-work-preserving systems reach saturation much faster than work-preserving counterparts as the arrival rate of secondary packets increases, particularly for larger packet sizes.
Conclusion and Future Work
This research provides a comprehensive statistical characterization of EDT in cognitive radio networks. While the current analysis assumes fixed-size secondary packets and negligible PU return during sensing intervals in the imperfect model, it establishes a foundation for more complex scenarios. Future work may explore multi-channel environments where SUs can switch frequencies to avoid waiting, as well as non-Poisson traffic models.
Summary updated