Abstract
We consider the problem of optimizing the freshness of status updates that are sent from a large number of low-power sources to a common access point. The source nodes utilize carrier sensing to reduce collisions and adopt an asynchronized sleep-wake scheduling strategy to achieve a target network lifetime (e.g., 10 years). We use age of information (AoI) to measure the freshness of status updates, and design sleep-wake parameters for minimizing the weighted-sum peak AoI of the sources, subject to per-source battery lifetime constraints. When the sensing time (i.e., the time duration of carrier sensing) is zero, this sleep-wake design problem can be solved by resorting to a two-layer nested convex optimization procedure; however, for positive sensing times, the problem is non-convex. We devise a low-complexity solution to solve this problem and prove that, for practical sensing times that are short, the solution is within a small gap from the optimum AoI performance. When the mean transmission time of status-update packets is unknown, we devise a reinforcement learning algorithm that adaptively performs the following two tasks in an “efficient way”: a) it learns the unknown parameter, b) it also generates efficient controls that make channel access decisions. We analyze its performance by quantifying its “regret”, i.e., the sub-optimality gap between its average performance and the average performance of a controller that knows the mean transmission time. Our numerical and NS-3 simulation results show that our solution can indeed elongate the batteries lifetime of information sources, while providing a competitive AoI performance.
Introduction
In massive Internet of Things (IoT) networks, maintaining information freshness while ensuring extreme energy efficiency is a critical challenge. Applications such as environmental monitoring or autonomous coordination require timely updates, yet sensors often operate in remote areas where battery replacement is impractical. Traditional scheduling focused on throughput or delay often fails to minimize the Age of Information (AoI), a metric specifically designed to quantify data freshness. Furthermore, centralized synchronization for thousands of nodes introduces prohibitive overhead. This paper proposes an asynchronous sleep-wake scheduling framework that minimizes weighted-sum peak AoI under strict multi-year battery lifetime constraints.
Age of Information and System Model
The freshness of status updates from a source $l$ is measured by the AoI, denoted as $\Delta_l(t)$. As shown in Figure 1, the AoI increases linearly over time and drops to a lower value upon the successful reception of a new update at the Access Point (AP). The peak AoI, $\Delta_{l,i}^{\text{peak}}$, represents the maximum age reached just before the $i$-th update is received.
The system consists of $M$ sources contending for a single AP. Each source $l$ has a power budget that must sustain it for a target lifetime (e.g., 10–15 years).
Asynchronous Sleep-Wake Scheduling
To avoid the overhead of global synchronization, the authors propose an asynchronized protocol. A source node wakes up and performs carrier sensing for a duration $t_s$. If the channel is idle, it transmits its packet; if busy, it immediately returns to sleep. This mechanism is illustrated in Figure 2, where multiple sources contend for the channel.
This approach is highly scalable for dense networks. The protocol handles collisions and busy channels by reverting to sleep mode, thereby conserving energy while maintaining an independent schedule for each node.
Optimization and Learning
The core of the research is a non-convex optimization problem aimed at minimizing the weighted average peak AoI. The authors identify two regimes: energy-adequate, where nodes can potentially saturate the channel, and energy-scarce, where nodes must sleep extensively to meet lifetime targets.
Despite the non-convexity introduced by non-zero sensing times ($t_s > 0$), a low-complexity, near-optimal closed-form solution is derived. The authors prove that the optimality gap vanishes as $t_s$ becomes small relative to the mean transmission time $E[T]$.
For scenarios where $E[T]$ is unknown, a Certainty Equivalence (CE) reinforcement learning algorithm is implemented. The AP estimates network parameters in real-time and broadcasts updated scheduling parameters, achieving $o(H)$ regret and ensuring long-term convergence to optimal performance without the high computational complexity of more optimistic RL variants.
Key Results and Scalability
Evaluation through Matlab and high-fidelity NS-3 simulations (IEEE 802.11b environment) yielded several key findings:
- Battery Longevity: The scheduler can maintain a 25-year battery life while keeping average update age to approximately 12 minutes in dense networks.
- Near-Optimality: For practical sensing times (e.g., 40µs), the solution is within 2% of the theoretical optimum in NS-3 simulations.
- Scalability: Performance remains robust as the number of sources increases to $10^5$, with peak age increasing only linearly.
- Throughput vs. Freshness: The study confirms that throughput-optimal schedulers are suboptimal for AoI, as they often lead to congestion and stale data; AoI requires more precise timing of updates.
Conclusion
This work provides a practical framework for massive IoT deployments, balancing the need for fresh data with extreme energy constraints. By leveraging asynchronous random access and low-complexity optimization, the proposed solution offers a scalable and efficient path toward long-lived, responsive sensor networks. Future work may address hidden terminal problems in multi-hop environments and dynamic priority weighting for critical events.
Summary updated