In real-time Industrial Internet of Things (IIoT), e.g., monitoring and control scenarios, the freshness of data is crucial to maintain the system functionality and stability. In this paper, we propose an AoI-Aware Deep Learning (AA-DL) approach to minimize the Peak Age of Information (PAoI) in D2D-assisted IIoT networks. Particularly, we analyzed the success probability and the average PAoI via stochastic geometry, and formulate an optimization problem with the objective to find the optimal scheduling policy that minimizes PAoI. In order to solve the non-convex scheduling problem, we develop a Neural Network (NN) structure that exploits the Geographic Location Information (GLI) along with feedback stages to perform unsupervised learning over randomly deployed networks. Our motivation is based on the observation that in various transmission contexts, the wireless channel intensity is mainly influenced by distancedependant path loss, which could be calculated using the GLI of each link. The performance of the AA-DL method is evaluated via numerical results that demonstrate the effectiveness of our proposed method to improve the PAoI performance compared to a recent benchmark while maintains lower complexity against the conventional iterative optimization method.

翻译：在实时工业物联网（IIoT）场景中，例如监控与控制应用，数据的时效性对维持系统功能与稳定性至关重要。本文提出了一种AoI感知深度学习方法（AA-DL），旨在最小化D2D辅助IIoT网络中的峰值信息年龄（PAoI）。具体而言，我们通过随机几何分析成功概率与平均PAoI，并构建了一个优化问题，其目标为寻找能最小化PAoI的最优调度策略。为求解这一非凸调度问题，我们设计了一种神经网络（NN）结构，该结构利用地理位置信息（GLI）结合反馈阶段，在随机部署网络上进行无监督学习。我们的动机源于观察：在不同传输场景中，无线信道强度主要受距离相关的路径损耗影响，而路径损耗可通过每条链路的GLI计算。数值结果验证了AA-DL方法的性能，表明与近期基准方法相比，该方法能有效提升PAoI性能，同时相较于传统迭代优化方法保持较低复杂度。