Edge computing has become a key technology to enable various Internet of Things (IoT) applications. We envision that in the future, edge computing systems will be platforms like crowdsourcing websites where IoT application providers can offload their tasks to the platform and pay it via the edge computing services, and the platform needs to seek appropriate edge servers (ESs) for executing tasks. To ensure ESs participate and provide good performance, i.e., to meet the stringent QoS requirement of IoT applications, the platform needs to properly reward them for computing services. In this paper, we consider the scenario where both the platform and ESs are self-interested entities, and focus on the setting that the payment made by application providers is ex-ante. We formulate a joint revenue maximization, resource allocation and task offloading optimization problem, and tackle it with a Stackelberg game formulation. A centralized algorithm based on Bayesian Optimization is proposed to solve the game. Moreover, we consider a practical setting where the platform is unable to collect full information when application providers and ESs may refuse to provide their sensitive information. To this end, we develop a decentralized solution based on neural network optimization together with a privacy-preserving information exchange protocol. We evaluate both mechanisms through numerical studies, and results indicate they are effective as compared to representative baselines.

To address the challenge of secure federated modeling in fault diagnosis under imbalanced data scenarios for multi-railway high-speed train bogies, this study proposes a multi-rank federated distillation (MFD) framework aimed at enhancing the generalization capacity of clients with limited sample sizes. First, the MFD framework is designed to perform multiple distillation tasks, with each task’s loss function decoupled into two components to balance losses between target and non-target classes. Second, an adaptive weight adjustment strategy is introduced to efficiently train models by coordinating the loss outputs across these tasks. Third, to mitigate the learning costs associated with the MFD, clients share a foundational shallow network via model transfer while incorporating personalized modules to improve adaptability. By validating the proposed framework on datasets from high-speed train bogies across multiple railways, this study demonstrates its effectiveness in addressing challenges associated with secure federated modeling while maintaining satisfactory diagnostic performance. The findings present a viable solution for implementing federated learning among clients with imbalanced data in industrial applications.

To provide reliable and elastic Multi-access edge computing services, one feasible solution is to federate geographically proximate edge servers to form a logically centralized resource pool. Optimization of such systems, however, becomes challenging. In this paper, we study the problem of maximizing users’ QoE in a MEC-based network, through jointly optimizing service caching, resource allocation and task offloading decisions. We formulate a mixed-integer nonlinear programming (MINLP) problem for the task and establish its NP-hardness. To tackle it efficiently, we propose a novel two-stage algorithmic solution based on approximation and decomposition theory. The proposed algorithm achieves high system performance while at the same time, ensures all constraints from different layers are satisfied. Meanwhile, the structure of the algorithm also fits the multi-layer optimizing feature, making it suitable to be implemented at different layers. In addition, we propose a distributed and online version of our mechanism with very limited information exchange between MEC servers, and further demonstrate how the cost of service switches from real MEC systems can be incorporated into our framework. We evaluate our mechanisms through simulations with both synthetic and real-world traces, and results indicate they are effective as compared to representative baseline algorithms.

Due to device operating environment limitations and data privacy protection, it is frequently difficult to obtain sufficient high-quality labeled data from devices, resulting in an insufficient generalization ability of fault diagnosis model. Therefore, a high-performance federated learning framework is proposed in this work, which makes improvements in the procedure of model aggregation and local model training. In the model aggregation of central server, an optimization aggregation strategy in which forgetting Kalman filter (FKF) is combined with cubic exponential smoothing (CES) is proposed to improve the efficiency of federated learning. In the local model training of multiclient, a deep learning network combined with multiscale convolution, attention mechanism, and multistage residual connection is proposed, which is able to fully extract multiclient data features simultaneously. Meanwhile, experiments on two machinery fault datasets show that the proposed framework is capable of achieving high accuracy and strong generalization of fault diagnosis on the premise of protecting data privacy in actual industrial situations.

Estimating the target position of low-frequency sound sources in a shallow sea environment is difficult due to the high cost of hydrophone placement and the complexity of the propagation model. We propose a compressed recurrent neural network (C-RNN) model that compresses the signal received by a vector hydrophone into a dynamic sound intensity signal and compresses the target position of the sound source into a GeoHash code. Two types of data are used to carry out prior training on the recurrent neural network, and the trained network is subsequently used to estimate the target position of the sound source. Compared with traditional mathematical models, the C-RNN model functions independently under the complex sound field environment and terrain conditions, and allows for real-time positioning of the sound source under low-parameter operating conditions. Experimental results show that the average error of the model is 56 m for estimating the target position of a low-frequency sound source in a shallow sea environment.