Capsule Network (CapsNet) has demonstrated significant potential in visual recognition by capturing spatial relationships and part-whole hierarchies for learning equivariant feature representations. However, existing CapsNet and variants often rely on a single high-level feature map, overlooking the rich complementary information provided by multi-scale features. Furthermore, conventional feature fusion strategies, such as addition and concatenation, struggle to reconcile multi-scale feature discrepancies, leading to suboptimal classification performance. To address these limitations, we propose the Multi-Scale Patchify Capsule Network (MSPCaps), a novel architecture that integrates multi-scale feature learning and efficient capsule routing. Specifically, MSPCaps consists of three key components: a Multi-Scale ResNet Backbone (MSRB), a Patchify Capsule Layer (PatchifyCaps), and a Cross-Agreement Routing (CAR) block. First, the MSRB extracts diverse multi-scale feature representations from input images, preserving both fine-grained details and global contextual information. Second, the PatchifyCaps partitions these multi-scale features into primary capsules using a uniform patch size, equipping the model with the ability to learn from diverse receptive fields. Finally, the CAR block adaptively routes the multi-scale capsules by identifying cross-scale prediction pairs with maximum agreement. Unlike the simple concatenation of multiple self-routing blocks, CAR ensures that only the most coherent capsules (best part-to-whole pairs) contribute to the final voting. Our proposed MSPCaps achieves remarkable scalability and superior robustness, consistently surpassing multiple baseline methods in terms of classification accuracy, with configurations ranging from a highly efficient Tiny model (344.3K parameters) to a powerful Large model (10.9M parameters), highlighting its potential in advancing feature representation learning.

The recent advancements in large language models (LLMs) have significantly enhanced language understanding and content generation capabilities. However, the deployment of LLMs on resource-constrained Internet of Things (IoT) devices remains challenging due to their substantial computational and storage requirements. To address this issue, we propose a novel LLM pruning method, termed structurally-aware adaptive pruning (SAAP), to reduce computational and storage costs for LLMs while maintaining model performance. Specifically, SAAP first leverages maximum likelihood estimation to calibrate traditional structural importance metrics for LLM pruning. Next, it employs a Bayesian fusion approach to address the predictive uncertainty in multi-granularity metrics, enabling accurate assessments of structural importance for LLMs. Then, SAAP introduces a cross-layer importance alignment mechanism based on quantile mapping, which normalizes layer-wise importance scores to ensure consistent pruning from a global perspective. Furthermore, SAAP develops an efficient block-wise fine-tuning strategy for enhancing the performance of the LLM after pruning. To validate the effectiveness of SAAP, we conduct extensive experiments on nine open-source LLMs across two representative tasks—language modeling and zero-shot classification. Experimental results show that SAAP consistently outperforms several baseline methods, achieving accuracy improvements of 2.5%, 2.63%, and 2.44% on LLaMA-7B, Vicuna-7B, and LLaMA-13B when the pruning ratio is 50%. Finally, SAAP is implemented on a testbed—NVIDIA Jetson AGX Orin 32GB Developer Kit. Test results demonstrate that compared to the foundation LLM, SAAP enhances the inference speed by 86.86% at a pruning ratio of 50%, highlighting its potential for practical deployment on resource-constrained IoT devices.

Unmanned aerial vehicle (UAV)-enabled integrated sensing and communication (ISAC) is regarded as a key enabler for next-generation wireless systems. However, conventional fixed-position antennas limit the ability of UAVs to fully exploit their inherent potential. To overcome this limitation, we propose a UAV-enabled ISAC framework equipped with fluid antennas (FAs), where the mobility of antenna elements introduces additional spatial degrees of freedom to simultaneously enhance communication and sensing performance. A multi-objective optimization problem is formulated to maximize the communication rates of multiple users while minimizing the Cram´er-Rao bound (CRB) for the angle estimation of a single target. Due to excessively frequent updates of FA positions may lead to response delay, a three-timescale optimization framework is developed to jointly optimize transmit beamforming, FA positions, and UAV trajectory based on their characteristics. To solve the non-convexity of the problem, an alternating optimization-based algorithm is developed to obtain a sub-optimal solution. Numerical results show that the proposed scheme significantly outperforms various benchmark schemes, validating the effectiveness of integrating the FA technology into the UAV-enabled ISAC systems.

Driven by the great success of Large Language Models (LLMs) in the 2D image domain, their application in 3D scene understanding has emerged as a new trend. A key difference between 3D and 2D is that the situation of an egocentric observer in 3D scenes can change, resulting in different descriptions (e.g., "left" or "right"). However, current LLM-based methods overlook the egocentric perspective and use datasets from a global viewpoint. To address this issue, we propose a novel approach to automatically generate a situation-aware dataset by leveraging the scanning trajectory during data collection and utilizing Vision-Language Models (VLMs) to produce high-quality captions and question-answer pairs. Furthermore, we introduce a situation grounding module to explicitly predict the position and orientation of the observer’s viewpoint, thereby enabling LLMs to ground situation descriptions in 3D scenes. We evaluate our approach on several benchmarks, demonstrating that our method effectively enhances the 3D situational awareness of LLMs while significantly expanding existing datasets and reducing manual effort.

Large Language Models (LLMs) exhibit remarkable capabilities in the hierarchical decomposition of complex tasks through semantic reasoning. However, their application in embodied systems faces challenges in ensuring reliable execution of subtask sequences and achieving one-shot success in long-term task completion. To address these limitations in dynamic environments, we propose Closed-Loop Embodied Agent (CLEA)—a novel architecture incorporating four specialized open-source LLMs with functional decoupling for closed-loop task management. The framework features two core innovations: (1) Interactive task planner that dynamically generates executable subtasks based on the environmental memory, and (2) Multimodal execution critic employing an evaluation framework to conduct a probabilistic assessment of action feasibility, triggering hierarchical re-planning mechanisms when environmental perturbations exceed preset thresholds. To validate CLEA’s effectiveness, we conduct experiments in a real environment with manipulable objects, using two heterogeneous robots for object search, manipulation, and search-manipulation integration tasks. Across 12 task trials, CLEA outperforms the baseline model, achieving a 67.3% improvement in success rate and a 52.8% increase in task completion rate. These results demonstrate that CLEA significantly enhances the robustness of task planning and execution in dynamic environments. Our code is available at https://sp4595.github.io/CLEA/.

Federated continual learning (FCL) has garnered increasing attention for its ability to support distributed computation in environments with evolving data distributions. However, the emergence of new tasks introduces both temporal and cross-client shifts, making catastrophic forgetting a critical challenge. Most existing works aggregate knowledge from clients into a global model, which may not enhance client performance since irrelevant knowledge could introduce interference, especially in heterogeneous scenarios. Additionally, directly applying decentralized approaches to FCL suffers from ineffective group formation caused by task changes. To address these challenges, we propose a decentralized dynamic cooperation framework for FCL, where clients establish dynamic cooperative learning coalitions to balance the acquisition of new knowledge and the retention of prior learning, thereby obtaining personalized models. To maximize model performance, each client engages in selective cooperation, dynamically allying with others who offer meaningful performance gains. This results in non-overlapping, variable coalitions at each stage of the task. Moreover, we use coalitional affinity game to simulate coalition relationships between clients. By assessing both client gradient coherence and model similarity, we quantify the client benefits derived from cooperation. We also propose a merge-blocking algorithm and a dynamic cooperative evolution algorithm to achieve cooperative and dynamic equilibrium. Comprehensive experiments demonstrate the superiority of our method compared to various baselines. Code is available at: https://github.com/ydn3229/DCFCL.

Federated learning (FL) has become a transformative paradigm for distributed machine learning over wireless networks. However, the performance of FL is hindered by the unreliable communication links between resource-constrained Internet of Things (IoT) devices and the central server. To overcome this challenge, we propose a novel framework that employs an unmanned aerial vehicle (UAV) as a mobile server to enhance the FL training process. By capitalizing on the UAV’s mobility, we establish strong line-of-sight connections with IoT devices, thereby enhancing communication reliability and capacity. To maximize training efficiency, we formulate a latency minimization problem that jointly optimizes bandwidth allocation, computing resources, transmit power for both the UAV and IoT devices, and the flight trajectory of the UAV. Subsequently, we analyze the required rounds of the IoT devices training and the UAV aggregation for FL convergence. Based on the convergence constraint, we transform the problem into three subproblems and develop an efficient alternating optimization algorithm to solve this problem. Additionally, we provide a thorough analysis of the algorithm’s convergence and computational complexity. Extensive numerical results demonstrate that the proposed algorithm-based scheme not only surpasses existing benchmark schemes in reducing latency up to 15.29%, but also achieves training efficiency that nearly matches the ideal scenario.

3D Visual Grounding (3DVG) aims at localizing 3D object based on textual descriptions. Conventional supervised methods for 3DVG often necessitate extensive annotations and a predefined vocabulary, which can be restrictive. To address this issue, we propose a novel visual programming approach for zero-shot open-vocabulary 3DVG, leveraging the capabilities of large language models (LLMs). Our approach begins with a unique dialog-based method, engaging with LLMs to establish a foundational understanding of zero-shot 3DVG. Building on this, we design a visual program that consists of three types of modules, i.e., view-independent, view-dependent, and functional modules. These modules, specifically tailored for 3D scenarios, work collaboratively to perform complex reasoning and inference. Furthermore, we develop an innovative language-object correlation module to extend the scope of existing 3D object detectors into open-vocabulary scenarios. Extensive experiments demonstrate that our zero-shot approach can outperform some supervised baselines, marking a significant stride towards effective 3DVG. Code is available at https://curryyuan.github.io/Z5VG3D.

This article considers a heterogeneous mobile-edge computing (HMEC) system with multiple energy-limited Internet of Things (IoT) devices and an unmanned aerial vehicle (UAV). The UAV can supply energy to all the IoT devices through wireless power transfer. To maximize the utilization of the communication and computation resources, all the IoT devices are divided into two groups, i.e., the active devices and the idle devices. The UAV and the idle devices assist the active devices in executing computing tasks. We formulate an optimization problem that maximizes the minimum task computation data volume among all the active devices by jointly optimizing the UAV trajectory and the communication and computation resource allocation. Since the problem is nonconvex, we decompose the problem into two subproblems: 1) the UAV trajectory design and the computation resource allocation and 2) the time allocation. We utilize a block coordinate descent approach to solve these two subproblems alternately. Simulation results demonstrate that the proposed algorithm can provide an optimized trajectory robust to different initializations. Additionally, compared to the benchmark algorithms, our proposed algorithm shows superior performance in terms of system efficiency and computation data volume.

Signal detection and modulation classification are two crucial tasks in various wireless communication systems. Different from prior works that investigate them independently, this paper studies the joint signal detection and automatic modulation classification (AMC) by considering a realistic and complex scenario, in which multiple signals with different modulation schemes coexist at different carrier frequencies. We first generate a coexisting RADIOML dataset (CRML23) to facilitate the joint design. Different from the publicly available AMC dataset, ignoring the signal detection step and containing only one signal, our synthetic dataset covers the more realistic multiple-signal coexisting scenario. Then, we present a joint framework for detection and classification (JDM) for such a multiple-signal coexisting environment, which consists of two modules for signal detection and AMC, respectively. In particular, these two modules are interconnected using a designated data structure called “proposal”. Finally, we conduct extensive simulations over the newly developed dataset, which demonstrate the effectiveness of our designs. Our code and dataset are now available as open-source resources athttps://github.com/Singingkettle/ChangShuoRadioData.

The deployment of generative adversarial networks (GANs) in wireless networks faces three key challenges of limited devices’ computational capability, scarce communication resources, and severe data privacy leakage. To address these issues, this paper proposes a new distributed framework for training GANs based on ensemble learning. First, multiple discriminators are trained at many devices using their local datasets. A generator is then trained at a central server by aggregating devices’ discriminators in an ensemble manner. The per-round training time is established. Finally, simulation results show that the proposed framework can simultaneously reduce the training time and improve the learning performance as compared with an existing framework.

Training task in classical machine learning models, such as deep neural networks, is generally implemented at a remote cloud center for centralized learning, which is typically time-consuming and resource-hungry. It also incurs serious privacy issue and long communication latency since a large amount of data are transmitted to the centralized node. To overcome these shortcomings, we consider a newly-emerged framework, namely federated edge learning, to aggregate local learning updates at the network edge in lieu of users' raw data. Aiming at accelerating the training process, we first define a novel performance evaluation criterion, called learning efficiency. We then formulate a training acceleration optimization problem in the CPU scenario, where each user device is equipped with CPU. The closed-form expressions for joint batchsize selection and communication resource allocation are developed and some insightful results are highlighted. Further, we extend our learning framework to the GPU scenario. The optimal solution in this scenario is manifested to have the similar structure as that of the CPU scenario, recommending that our proposed algorithm is applicable in more general systems. Finally, extensive experiments validate the theoretical analysis and demonstrate that the proposed algorithm can reduce the training time and improve the learning accuracy simultaneously.

The implementation of artificial intelligence (AI) in wireless networks is becoming more and more popular because of the growing number of mobile devices and the availability of huge amount of data. Directly transmitting data for centralized learning will cause long communication latency and may incur severe privacy issue as well. To address these issues, we consider the importance-aware federated edge learning (FEEL) system in this paper. Based on the relation between loss decay and gradient norm, a learning efficiency maximization problem is formulated by jointly considering the communication resource allocation and data selection. The closed-form results for optimal communication resource allocation and data selection are both developed, where some insights are also highlighted. Finally, the test results show that the proposed algorithm can effectively reduce the training latency and improve the learning accuracy as compared with some benchmark algorithms.

With the combination of deep learning and mobile edge computing, the accuracy of the computer vision task in mobile augmented reality (AR) applications can be significantly improved along with the enhancement on the end-to-end latency and energy efficiency. However, no architecture-based delay model for convolutional neural networks (CNNs) has been proposed in edge computing. In this paper, we first develop a new delay model to characterize the relation between the processing delay and the input image size of general CNN models. Then, we formulate a non-convex optimization problem to maximize the learning accuracy under the communication and computation resource constraints. By problem transformation, the optimal resource allocation policy is derived in closed-form and low-complexity search algorithm is also developed. Finally, test results validate the applicability of the delay model and demonstrate the learning accuracy improvement of the proposed algorithm.

It is an essential goal for future wireless networks to provide better artificial intelligent services. In this paper, we investigate the joint communication and computation resource optimization in the mobile edge learning system to support augmented reality applications, where the convolutional neural networks (CNNs) are deployed at the edge server. For such a system, we first develop a delay model to characterize the relation between the computation latency and the input image size of general CNN models. Then, we formulate a mixed integer nonlinear optimization problem to maximize the system computation capacity under the constraints of learning accuracy, end-to-end latency, and energy consumption. To solve this problem, we first investigate maximizing the system learning accuracy under the communication and computation resource constraints. The optimal resource allocation policy can be achieved by a low-complexity search algorithm. We further prove that the original problem is NP-hard and propose an efficient heuristic algorithm with a newly-developed offloading priority function. An upper bound for the proposed algorithm is also derived. Finally, test results validate the applicability of the delay model and demonstrate the performance improvement of the proposed algorithm as compared with the existing algorithms.

In cellular federated edge learning (FEEL), multiple edge devices holding local data jointly train a neural network by communicating learning updates with an access point without exchanging their data samples. With very limited communication resources, it is beneficial to schedule the most informative local learning updates. This paper focuses on FEEL with gradient averaging over participating devices in each round of communication. A novel scheduling policy is proposed to exploit both diversity in multiuser channels and diversity in the “importance” of the edge devices' learning updates. First, a new probabilistic scheduling framework is developed to yield unbiased update aggregation in FEEL. The importance of a local learning update is measured by its gradient divergence. If one edge device is scheduled in each communication round, the scheduling policy is derived in closed form to achieve the optimal trade-off between channel quality and update importance. The probabilistic scheduling framework is then extended to allow scheduling multiple edge devices in each communication round. Numerical results obtained using popular models and learning datasets demonstrate that the proposed scheduling policy can achieve faster model convergence and higher learning accuracy than conventional scheduling policies that only exploit a single type of diversity.

The mobile edge computing (MEC) and device-to-device (D2D) communications take advantage of the proximity for supporting high-speed mobile computing and high-rate data communications, respectively. In this paper, we integrate both techniques to further improve the computation capacity of the cellular networks by proposing the D2D-MEC technique. We aim to maximize the number of supported devices and formulate a mixed integer non-linear problem. To solve it, we decouple it into two subproblems and prove that the optimal solutions to the two subproblems compose the optimal solution to the original problem. The first one minimizes the required edge computation resource for a given D2D pair while the second one maximizes the number of supported devices via optimal D2D pairing. Then, by solving two subproblems, the optimal algorithm is developed and some insightful results are also highlighted. Finally, numerical results show that combining D2D communications with MEC can significantly enhance the computation capacity of the system.

In this paper, we investigate the latency minimization resource allocation problem in a hierarchical cloud-edge coexistence system by optimally splitting tasks for partial cloud computing and partial edge computing. A joint communication and computation resource allocation problem is first formulated and the structural characteristics are further analyzed. Next, by defining two novel parameters: the normalized backhaul communication capacity and the normalized cloud computation capacity, an optimal task splitting strategy is developed. With the help of these definitions, the joint communication and computation resource allocation policy can be devised in closed-form. Finally, numerical results demonstrate that the proposed collaborative cloud-edge computing scheme performs better than some baseline schemes in terms of minimizing the end-to-end latency of mobile devices.

The future 5G wireless networks aim to support high-rate data communications and high-speed mobile computing. To achieve this goal, the mobile edge computing (MEC) and device-to-device (D2D) communications have been recently developed, both of which take advantage of the proximity for better performance. In this paper, we integrate the D2D communications with MEC to further improve the computation capacity of the cellular networks, where the task of each device can be offloaded to an edge node and a nearby D2D device. We aim to maximize the number of devices supported by the cellular networks with the constraints of both communication and computation resources. The optimization problem is formulated as a mixed integer non-linear problem, which is not easy to solve in general. To tackle it, we decouple it into two subproblems. The first one minimizes the required edge computation resource for a given D2D pair, while the second one maximizes the number of supported devices via optimal D2D pairing. We prove that the optimal solutions to the two subproblems compose the optimal solution to the original problem. Then, the optimal algorithm to the original problem is developed by solving two subproblems, and some insightful results, such as the optimal transmit power allocation and the task offloading strategy, are also highlighted. Our proposal is finally tested by extensive numerical simulation results, which demonstrate that combining D2D communications with MEC can significantly enhance the computation capacity of the system.

Mobile edge computing (MEC) has become a dominant technology in the upcoming era of the 5th generation mobile networks. By offloading tasks from mobile devices to edge clouds provided by cellular base stations, both energy consumption and end-to-end delay of mobile tasks can be reduced. In this paper, we aim to optimize the latency performance of TDMA-based MEC systems by joint allocation of computation and communication resource. Our goal is to minimize the maximal delay of all devices in the system. We first simplify the optimization problem and convert it into a convex one. Then we derive the closed-form expression for the optimal resource allocation strategy and investigate the relationship between uplink and downlink resource allocation. A subgradient algorithm is also developed to solve the joint resource allocation problem. Finally, numerical simulation results are shown to verify that our proposal can achieve a better performance compared with the traditional schemes.

In this paper, we investigate the latency minimization resource allocation for a multi-user augmented reality (AR) system based on mobile edge computing (MEC). First, we develop a novel data sharing model for the delay-sensitive AR tasks. Then, by integrating the partial offloading scheme into the task processing, we formulate a weighted-sum latency minimization problem to improve the quality of experience (QoE) for AR devices. Both the optimal task segmentation strategy and the optimal joint resource allocation are derived in closed-form. Finally, numerical results show that the proposed partial task offloading with data sharing scheme can achieve a better delay performance as compared against some benchmark schemes.

By offloading intensive computation tasks to the edge cloud located at the cellular base stations, mobile-edge computation offloading (MECO) has been regarded as a promising means to accomplish the ambitious millisecond-scale end-to-end latency requirement of fifth-generation networks. In this paper, we investigate the latency-minimization problem in a multi-user time-division multiple access MECO system with joint communication and computation resource allocation. Three different computation models are studied, i.e., local compression, edge cloud compression, and partial compression offloading. First, closed-form expressions of optimal resource allocation and minimum system delay for both local and edge cloud compression models are derived. Then, for the partial compression offloading model, we formulate a piecewise optimization problem and prove that the optimal data segmentation strategy has a piecewise structure. Based on this result, an optimal joint communication and computation resource allocation algorithm is developed. To gain more insights, we also analyze a specific scenario where communication resource is adequate while computation resource is limited. In this special case, the closed-form solution of the piecewise optimization problem can be derived. Our proposed algorithms are finally verified by numerical results, which show that the novel partial compression offloading model can significantly reduce the end-to-end latency.

In this paper, we consider latency-minimization resource allocation for a multi-user mobile edge computation offloading (MECO) system. First, we develop a novel partial computation offloading model and then formulate the weighted-sum latency-minimization problem by optimally allocating the communication and computation resources. After that, the closed-form expression for the optimal data segmentation strategy is derived. Based on this result, we transform the original problem into a piecewise convex optimization problem and propose a sub-gradient algorithm to find the optimal resource allocation solution. Moreover, we analyze a specific scenario where communication resource is adequate while computation resource is limited. In this special case, the closed-form solution is devised. Finally, numerical results show that the partial computation offloading model can achieve a better performance than other two baseline schemes.

By offloading computationally intensive tasks to the edge cloud provided by the cellular base stations, the mobile edge computing (MEC) technique has the potential to realize the critical millisecond-scale latency requirement of next generation mobile services. In this paper, we investigate the joint subcarrier and power allocation problem in an orthogonal frequency division multiple access (OFDMA) based MEC system to minimize the maximal delay of each mobile device. The partial data offloading scenario is considered where mobile data can be computed at both local devices and the edge cloud. Since the problem is a combinatorial optimization one, we first propose a lower-bound algorithm by relaxing the channel allocation indicators into continuous variables. Then, to find a low-complexity feasible solution, we further develop a heuristic algorithm which separates subcarrier assignment and power allocation. The performances of our proposed algorithms are finally tested by extensive numerical simulations.