In this paper, we study a system in which a sensor forwards status updates to a receiver through an error-prone channel, while the receiver sends the transmission results back to the sensor via a reliable channel. Both channels are subject to random delays. To evaluate the timeliness of the status information at the receiver, we use the Age of Information (AoI) metric. The objective is to design a sampling policy that minimizes the expected time-average AoI, even when the channel statistics (e.g., delay distributions) are unknown. We first review the threshold structure of the optimal offline policy under known channel statistics and then reformulate the design of the online algorithm as a stochastic approximation problem. We propose a Robbins-Monro algorithm to solve this problem and demonstrate that the optimal threshold can be approximated almost surely. Moreover, we prove that the cumulative AoI regret of the online algorithm increases with rate$\mathcal {O}(\ln K)$, where$K$is the number of successful transmissions. In addition, our algorithm is shown to be minimax order optimal, in the sense that for any online learning algorithm, the cumulative AoI regret up to the$K$-th successful transmissions grows with the rate at least$\Omega (\ln K)$in the worst case delay distribution. Finally, we improve the stability of the proposed online learning algorithm through a momentum-based stochastic gradient descent algorithm. Simulation results validate the performance of our proposed algorithm.

Vehicular fog computing (VFC) offers a promising paradigm to alleviate vehicular computing burdens and ensure timely processing of computing tasks. In this paper, a heterogeneous VFC system leveraging hybrid visible light communication (VLC) and radio frequency (RF) communications is investigated, exploiting the interference resilience of VLC and the extended coverage of RF for efficient task offloading to multiple idle vehicles while harnessing their computing resources for parallel task computing. On this basis, an average task processing delay (TPD) minimization problem is formulated, thereby encompassing task offloading, computing and communication resource allocation for rapid processing of computing tasks. Then, the non-convex problem is decomposed into three subproblems and iteratively solved within a block coordinate descent (BCD) framework. Within this framework, the residual additive majorization-minimization algorithm is employed for task offloading and computing resource allocation, the transformed majorization-minimization method is applied for joint power adjustment of both VLC and RF, and the relaxed optimization and reconstruction approach is developed for subchannel assignment. Comprehensive simulations validate the superiority of the heterogeneous VFC system over local computing and VFC systems relying on VLC or RF, and demonstrate the convergence of the proposed BCD-based algorithm and its superiority over baselines. Additionally, the influence of key parameters on the average TPD is also characterized, providing essential design guidelines for practical implementations.

In this paper, a novel integrated control and communication (ICC) framework is proposed for vehicle platoons, ensuring platoon stability while optimizing energy consumption for communications. To address the high signaling overhead and long delays caused by multi-link competition in radio-frequency-based vehicle-to-vehicle communications, visible light communications (VLCs) are employed, allowing adjacent vehicles to establish collision-free links and simultaneously transmit motion status information (MSI) for platoon control. Besides, age of information (AoI) is utilized to quantify the heterogeneous timeliness of the MSI. Building upon this, a distributed linear control strategy is designed for vehicle platoons with a multi-predecessor-leader-multi-following information topology (IT) and heterogeneous AoIs. Afterward, through stability analysis, the maximum allowable AoI threshold is obtained to guarantee both internal stability and input-state string stability under external disturbances. Accordingly, taking the AoI threshold as a constraint, a long-term energy minimization problem is further formulated to optimize the energy consumed by multi-source, multi-hop, and multicast intra-platoon VLC while maintaining platoon stability. To solve this problem, an online distributed information scheduling policy based on Lyapunov optimization is proposed. Finally, simulation results reveal that the proposed ICC framework effectively maintains platoon stability across various ITs, while significantly reducing the energy consumption for communications to no more than 40% of that achieved by multiple baselines, thus highlighting its substantial potential in vehicle platoons.

The underwater Internet of Things (UIoT) is crucial in developing marine resources. However, due to the low data rate of underwater channels, it is difficult to have a central server to process data from numerous devices as using terrestrial communications. Therefore, decentralized federated learning (DFL) with communication-efficient modifications is a promising alternative to empower UIoT with artificial intelligence and collaborative training. However, existing DFL strategies rely on a carefully designed small aggregation weight when aggregating parameters from neighbor nodes to mitigate the compression error, resulting in a slow convergence rate. In addition, the effect of data compression under time-varying topologies is not considered in current DFL algorithms. In response to these problems, this work studies a DFL framework with underwater acoustic channel and time-varying topology. Firstly, considering the low data rate and dynamics of the acoustic channel, we propose a practical scheme for adaptive compression and device connectivity. Moreover, we combine data compression and the error-compensation technique with time-varying topology and propose a DFL algorithm with aggregation weights decaying over time to achieve fast convergence under non-independent and identically distributed (non-IID) data. We derive a convergence bound for the proposed algorithm with respect to compression and time-varying topology and demonstrate that it achieves the same asymptotic convergence rate as centralized FL with perfect communication. Simulation results show that, compared with DFL algorithms without decaying aggregation weights and centralized FL schemes, the proposed algorithm exhibits higher accuracy and faster convergence rate in underwater environments.

Among existing Neural Architecture Search methods, DARTS is known for its efficiency and simplicity. This approach applies continuous relaxation of network representation to construct a weight-sharing supernet and enables the identification of excellent subnets in just a few GPU days. However, performance collapse in DARTS results in deteriorating architectures filled with parameter-free operations and remains a great challenge to the robustness. To resolve this problem, we reveal that the fundamental reason is the biased estimation of the candidate importance in the search space through theoretical and experimental analysis, and more precisely select operations via information-based measurements. Furthermore, we demonstrate that the excessive concern over the supernet and inefficient utilization of data in bi-level optimization also account for suboptimal results. We adopt a more realistic objective focusing on the performance of subnets and simplify it with the help of the informationbased measurements. Finally, we explain theoretically why progressively shrinking the width of the supernet is necessary and reduce the approximation error of optimal weights in DARTS. Our proposed method, named IS-DARTS, comprehensively improves DARTS and resolves the aforementioned problems. Extensive experiments on NAS-Bench-201 and DARTS-based search space demonstrate the effectiveness of IS-DARTS.

The Autonomous Underwater Vehicle (AUV)- assisted Underwater Internet of Things (UIoT) has received much attention due to its potential to develop marine resources with big data analysis. Given the low data rates and distributed data, decentralized federated learning (DFL) emerges as a promising avenue, enabling artificial intelligence integration and collaborative training within the underwater environment. In this paper, we combine DFL with the underwater scenario for the first time. A novel DFL algorithm with decaying aggregation weight is proposed to achieve fast convergence rate under non-independent and identically distributed (non-IID) data. In addition, the DFL algorithm is tailored to underwater acoustic channels, and we integrate adaptive compression, device connectivity, and time-varying topology considerations to enable practical deployment. We provide convergence analysis under convexity and connectivity assumptions. Simulation experiments validate the performance using the MNIST dataset, highlighting its effectiveness for practical UIoT applications.

In this work, we study a system with a sensor forwarding status update to the receiver through an error-prone channel, and the receiver sends the transmission results to the sensor via a reliable link. We assume both transmission links suffer from random delays. We use Age of Information (AoI) to measure the freshness of the status information at the receiver. Our goal is to design a sampling policy that minimizes the expected time average AoI when the channel statistics are unknown. The problem is reformulated into a renewal-reward process optimization, and an online algorithm based on the Robbins-Monro algorithm is proposed. We prove that when the forward and backward transmission delays are bounded, the AoI difference between the online algorithm and the optimal policy decays with rate$\mathcal{O}(\ln K/K)$, where$K$is the number of successful transmissions. Simulation results validate the performance of our proposed algorithm.

Structural re-parameterization is a raising field, which aims at improving the performance of convolutional neural networks (CNNs) through training an over-parameterization model and transferring it into a compact inference model. However, the performance improvements of prior structural re-parameterization works often come at the cost of heavy extra training resources, which increases carbon emissions and limits the potential applications on large-scale industrial tasks. To this end, first, we conduct experiments with a series of blocks composed of multiple identical branches to investigate the mechanism behind the structural re-parameterization, and then provide an interpretation. Moreover, motivated by the studies of effective receptive fields in the biological visual systems and neural networks, we propose a novel compact block named circular mask block (CMB). Given a neural network, we replace the regular convolutional layer with CMB to construct a training architecture, which can be trained to gain an accuracy boost with No extra training parameters and limited extra training FLOPs. After training, the training architecture can be transformed into the original architecture for inference. Extensive experiments are performed on CIFAR-10 and ImageNet to evaluate the effectiveness of our method. For example, we improve 0.85% top-1 accuracy of ResNet-50 on ImageNet without extra training parameters and only 11.32M extra training FLOPs, which saves 434x training FLOPs compared with prior works.

Massive multiple-input multiple-output (MIMO) is one of the key techniques to achieve better spectrum and energy efficiency in 5G system. The channel state information (CSI) needs to be fed back from the user equipment to the base station in frequency division duplexing (FDD) mode. However, the overhead of the direct feedback is unacceptable due to the large antenna array in massive MIMO system. Recently, deep learning is widely adopted to the compressed CSI feedback task and proved to be effective. In this paper, a novel network named aggregated channel reconstruction network (ACRNet) is designed to boost the feedback performance with network aggregation and parametric rectified linear unit (PReLU) activation. The practical deployment of the feedback network in the communication system is also considered. Specifically, the elastic feedback scheme is proposed to flexibly adapt the network to meet different resource limitations. Besides, the network binarization technique is combined with the feature quantization for lightweight and practical deployment. Experiments show that the proposed ACRNet outperforms loads of previous state-of-the-art networks, providing a neat feedback solution with high performance, low cost and impressive flexibility.