Embedded Files
Energy-aware protocols
References
D. Ruíz-Guirola, P. Raghuwanshi, G. Jesus, M. Ashraf, O. López, "Context-awareness for Dependable Low-Power IoT", available at https://arxiv.org/pdf/2510.23125.
S. Jahanbazi, M. Ashraf, O. L´opez, "MDP-based Energy-aware Task Scheduling for Battery-less IoT", available at https://arxiv.org/pdf/2510.23820.
O. L´opez, M. Ashraf, S. Nasser, G. Jesus, R. Singh, M. Filippou, J. Famaey, "Foundations for energy-aware zero-energy devices: From energy sensing to adaptive protocols", https://arxiv.org/abs/2507.22740.
M. Ashraf, S. Jahanbazi, O. L´opez, "Evaluating Task Execution Performance Under Energy Measurement Overhead", available at https://arxiv.org/pdf/2508.08757.
Context-awareness for Dependable Low-Power IoT
Context-awareness for dependable IoT.
Energy-limited/EH-powered network.
We propose a context-aware approach to protocol design as a foundation for allowing dependable operation in low-power, energy-constrained IoT networks. We present an edge-centric control plane that performs a cross-layer mapping from context to dependability targets, deriving per-device policies. Additionally, we introduce application-specific awareness mechanisms that continuously configure system parameters based on contextual factors such as energy availability, network dynamics, and task requirements. Finally, through an assessment of representative use cases, we demonstrate significant improvements in detection latency, accuracy, and overall availability, relative to context-agnostic baselines, all achieved with minimal control-plane overhead.
The figures above compares system dependability in terms of availability, reliability, and scalability, under context-aware intelligent duty cycling and sensing against a baseline using standard blind duty cycling. Context here includes the spatial and time correlation between IoTDs and events, as well as the EI received at the EdN (i.e., EH, consumption profile, and energy storage state of the IoTDs). The proposed context- aware scheme combines RL with a K-nearest neighbor (KNN)-based model to dynamically adjust duty cycling and wake-up decisions. As shown in figure, the proposed mechanisms improve event detection reliability by up to 3 times while reducing the event reporting delay by up to 56%.
Context-aware duty cycling and sensing also increase the mean time to failure (MTTF), defined as the expected time to failure (i.e., the average operating time before the IoT system experiences a failure). Herein, we define MTTF as the point at which more than 10%, 25%, and 50% of the network area is covered by energy-depleted IoTDs. These 10/25/50% levels serve as operational tiers (e.g., minor degradation, impaired service, and critical failure), and model escalating mitigation actions. As illustrated in figure, our approach increases the MTTF by up to two orders of magnitude while illustrating improvements in availability. Our dependability assessment encompasses availability, reliability, resilience, energy sustainability, and event detection latency. Overall, the proposed context-aware approach maintains long-term system dependability through intelligent sensing and energy-efficient operation.
MDP-based Energy-aware Task Scheduling for Battery-less IoT
We consider a battery-less IoT device that must periodically report sensing measurements to a monitoring center. We adopt the Markov decision process (MDP) framework to handle energy variability while aiming to maximize the long-term task completion rate. For this, we first identify its com-ponents and then define two appropriate reward functions. We demonstrate the inherent properties associated with the MDP formulation and the related optimal policy. Subsequently, we solve the resulting optimization problem, leading to the optimal stationary threshold-based (OSTB) scheduling.
This figure depicts the task completion rate per main interval over time, obtained using the Monte Carlo method. Note that OSTB consistently achieves a higher rate of successful task execution, while ALAP allows more time for EH, which in turn increases energy availability and supports safer task execution. However, within the constraints of the considered system model, deferring task execution until the latest permissible moment can reduce the likelihood of timely completion. This limitation stems from the stochastic EH nature, where delaying the “sensing” task until its deadline may leave insufficient time or energy to complete the subsequent “transmitting” task within the remaining interval. This effect ultimately lowers the overall task completion rate. Moreover, it is important to note that the difference in rate of completed tasks between ALAP and OSTB becomes less pronounced when a larger capacitor is used. In this case, the voltage variation across the capacitor is more stable. Since the average directly depends on the capacitor voltage, the resulting average number also exhibits more stable behavior.
llustrates the task completion rate over a 2000-second duration for various Vout and different maximum values of the harvested current, modeled as uniform distribution. The results indicate that when the turn-off threshold Vout is set to 1.8V, the minimum γ required to achieve the maximum task completion rate (i.e., successful execution of both tasks within each main interval) is 7.3mA. In contrast, for higher threshold voltages of 2.1V and 2.4V, the corresponding minimum γ values increase to 11.2mA and exceed 12mA, respectively. Furthermore, it is shown that the task completion rates of both OSTB and ALAP approaches converge for higher values of γ. This observation implies that, as the harvested current increases, deferring task execution until the latest permissible time becomes a viable strategy, particularly in scenarios where execution delay is not a critical concern. These results illustrate the fact that the performance improvement is more pronounced when the support range of harvesting current (i.e., γ) is small.
Foundations for energy-aware zero-energy devices: From energy sensing to adaptive protocols
Zero-energy devices (ZEDs) are key enablers of sustainable Internet of Things networks by operating solely on harvested ambient energy. Their limited and dynamic energy budget calls for protocols that are energy-aware and intelligently adaptive. However, designing effective energy-aware protocols for ZEDs requires theoretical models that realistically reflect device constraints. Indeed, existing approaches often oversimplify key aspects such as energy information (EI) acquisition, tasklevel variability, and energy storage dynamics, limiting their practical relevance and transferability. This article addresses this gap by offering a structured overview of the key modeling components, trade-offs, and limitations involved in energy-aware ZED protocol design. For this, we dissect EI acquisition methods and costs, characterize core operational tasks, analyze energy usage models and storage constraints, and review representative protocol strategies. Moreover, we offer design insights and guidelines on how ZED operation protocols can leverage EI, often illustrated through selected in-house examples. Finally, we outline key research directions to inspire more efficient and scalable protocol solutions for future ZEDs.
As shown in the top figure on the left, the successful execution of the tasks under the energy-aware scheme depends on Ec for a given value of Q. Specifically, a higher Ec results in performance degradation when Q is small. Moreover, the size of the task buffer positively affects the performance of the energy-aware scheme, while the performance of the energyblind scheme remains unaffected by the task buffer’s size for larger values of F and decreases with B for smaller values of F.
From the bottom figure, and as expected, a higher EM is always beneficial. However, the potential gains are negligible for higher F and Q as the performance depends more on the task buffer size and the arrival rates in these regimes. In particular, if the task buffer becomes full, subsequent task arrivals will contribute to task failure even if enough energy is available.
Figures on the left show the average AoI as a function of the energy threshold δ used for the partially-aware approach (top) and the energy arrival rate p′ (bottom). Since the energy threshold is not applicable for the fully-aware case and the baseline, their corresponding results appear as straight lines. Observe that the partially-aware setup achieves minimum average AoI when the threshold equals EM and events are rarer, i.e., lower p, resulting in performance similar to that of the fully-aware approach. On the other hand, when events are more frequent, i.e., larger p, the partially-aware approach can outperform the fully-aware one if the threshold is properly tuned according to the system parameters. However, modifying the threshold may not be possible in simple ZED implementations, limiting their deployment dynamicity/scalability in practice. Meanwhile, a feedback link from the base station to the ZEDs may be required when threshold optimization is possible. All these aspects require further research.
Lightweight protocols
References
M. Ashraf, S. Jahanbazi, O. L´opez, "Evaluating Task Execution Performance Under Energy Measurement Overhead", available at https://arxiv.org/pdf/2508.08757.
Evaluating task execution performance under energy measurement overhead
This work highlights operational parameters, such as energy measurement frequency and task execution frequency, which can be tuned to improve the task execution performance of an EH-IoT device. To this end, we consider energy blind (EB) and energy-aware (EA) task decision approaches and compare their task completion rate performance. We show that, for specific hardware design parameters of an EH-IoT device, there exists an optimal energy measurement/task execution frequency that can maximize the task completion rate in both approaches. Moreover, if these parameters are not chosen appropriately, then energy measurement costs can cause EA scheduling to underperform compared to EB scheduling.
The above figures show the task completion rate performance of the EB and EA scheduling with respect to task arrival rate, respectively. It is observed that the performance of both schemes degrades with a task arrival rate increase. This is because a larger can arrival rate (p) over burden the system and cause task buffer to remain full for most time slots. This effect can be mitigated by increasing the task buffer size. However, the improvement brought in by the use of increased task buffer size are neutralized if task execution frequency(F) in EB scheduling and energy measurement frequency ( Q) in EA scheduling is increased. In EB scheduling, this is due to the fact that a higher will cause delays in consecutive task executions, which will result in its increased occupancy rate. This effect will cause any subsequently arriving tasks to be dropped. For EA scheduling, the decrease in performance is caused by the use of conservative energy estimates for all those time slots where energy measurements do not occur. Hence, a lower estimate for the available energy in those time slots causes task executions to be further delayed, thus, resulting in degradation of performance.
Page updated
Google Sites
Report abuse