Learning-based sensing and computing decision for data freshness in edge computing-enabled networks

TL;DR

This work tackles data freshness in edge-computing wireless sensor networks by introducing the η-coverage probability to capture both spatial and temporal information quality. It develops two SCD strategies: a probability-based method deriving closed-form or tractable expressions for optimal sensing and EC offloading, and an RL-based method using a multi-agent POMDP with centralized training (MADDPG) to adapt decisions in real time. The results show RL-SCD uniformly outperforms baseline approaches in both single pre-charged and energy-harvesting sensor scenarios, demonstrating improved η-coverage and more efficient energy use. The proposed framework provides practical guidance on applying joint sensing/computation decisions to maintain data freshness in dynamic EC-enabled networks, with implications for AI-driven edge applications.

Abstract

As the demand on artificial intelligence (AI)-based applications increases, the freshness of sensed data becomes crucial in the wireless sensor networks. Since those applications require a large amount of computation for processing the sensed data, it is essential to offload the computation load to the edge computing (EC) server. In this paper, we propose the sensing and computing decision (SCD) algorithms for data freshness in the EC-enabled wireless sensor networks. We define the η-coverage probability to show the probability of maintaining fresh data for more than η ratio of the network, where the spatial-temporal correlation of information is considered. We then propose the probability-based SCD for the single pre-charged sensor case with providing the optimal point after deriving the η-coverage probability. We also propose the reinforcement learning (RL)- based SCD by training the SCD policy of sensors for both the single pre-charged and multiple energy harvesting (EH) sensor cases, to make a real-time decision based on its observation. Our simulation results verify the performance of the proposed algorithms under various environment settings, and show that the RL-based SCD algorithm achieves higher performance compared to baseline algorithms for both the single pre-charged sensor and multiple EH sensor cases.

Figures (8)

• Figure 1: Network description of EC-enabled wireless sensor networks.
• Figure 2: Example path of $\Delta_{n}^{\text{rx}}(t)$ for EH model.
• Figure 3: Example of the AoI path of the single pre-charged sensor case, $\Delta^{\text{tx}}(t)$ and $\Delta^{\text{rx}}(t)$, from the $(k-1)$-th to the $(k+1)$-th successful update.
• Figure 4: Proposed RL-based SCD algorithm. For readability, the neural networks for only two sensors are shown in the figure.
• Figure 5: $\eta$-coverage probability for the single pre-charged sensor case.

Theorems & Definitions (3)

• Lemma 1: Expected Inter-Update Time
• Lemma 2: Expected Violation Time
• Remark 1