Abstract
In Internet of Things (IoT) networks, the timeliness of status updates is critical for system performance. This paper investigates the Age of Information (AoI) in a multicast setting where an Access Point (AP) transmits updates to multiple devices. We analyze the impact of imposing transmission deadlines—both fixed and random—to mitigate the "slowest receiver" problem inherent in multicast. By deriving closed-form expressions for average AoI, we demonstrate that an optimal deadline exists which minimizes the age of data at the receivers. Our findings show that dropping stale updates is a superior strategy for maintaining information freshness compared to persistent retransmission.
Introduction
The value of data in IoT applications, such as autonomous vehicle safety or real-time sensor monitoring, is often a function of its "freshness." Age of Information (AoI) has emerged as a key metric to quantify this freshness, defined as the time elapsed since the generation of the most recently received update. While unicast AoI is well-studied, multicast scenarios present a unique challenge: the service time is dictated by the slowest required receiver. If the system waits for every device to acknowledge a packet, the information may become stale for the entire network. This paper explores how "drop-dead" timers or deadlines can clear the path for newer, fresher updates.
System Model and Methodology
We consider a system where an AP multicasts time-stamped updates to $N$ IoT devices. A transmission is deemed successful once a subset of $K$ devices receives the update. We assume a "generate-at-will" source model, where a new update is generated immediately after the previous transmission cycle ends (either through success or deadline expiration).
To analyze this, we model the AoI evolution as a stochastic process. The AoI $\Delta_n(t)$ follows a sawtooth pattern, increasing linearly over time and dropping only when a new update is successfully delivered.
As illustrated in Figure 1, the introduction of a deadline limits the duration of any single transmission attempt ($Y_{n,q}$). If the deadline expires before $K$ devices receive the packet, the current update is discarded, and a fresh one is generated. We utilize order statistics to derive the probability density functions (PDFs) of service times, allowing for the calculation of the first and second moments of inter-generation times, which are essential for determining the average AoI.
Key Results and Discussion
Our analysis reveals several critical insights into the behavior of AoI in multicast systems, validated through Monte-Carlo simulations in MATLAB involving 100,000 consecutive status updates.
The Deadline Paradox and the "Goldilocks" Zone
Contrary to the intuition that constraints hinder performance, deadlines are shown to be strictly beneficial for AoI. There exists an optimal "Goldilocks" deadline value that minimizes the average age.
As shown in Figure 2, if the deadline $T_D$ is too short, the probability of successful delivery is low, leading to frequent drops and high AoI. Conversely, if the deadline is too long, the system spends excessive time trying to reach the $K$-th device, allowing the data to age. The U-shaped curves in Figure 2 clearly identify the optimal operating points for different service rates ($\lambda_s$).
Fixed vs. Random Deadlines
The results indicate that fixed deadlines generally provide a lower minimum average AoI than random deadlines because they effectively bound the worst-case variability in service times. However, random deadlines can be more effective in high-deadline regimes where the system is less constrained. Furthermore, systems with deadlines consistently outperform those without, especially as the service rate increases.
Conclusion
This work demonstrates that enforcing transmission deadlines is a powerful tool for optimizing information freshness in multicast IoT networks. By strategically dropping stale updates, systems can achieve a significantly lower average AoI. Future research will extend this model to stochastic arrival processes and heterogeneous network environments.
Summary updated