Shared-nothing geo-distributed SQL databases, such as CockroachDB, are increasingly vital for enterprise applications requiring data resilience and locality. However, we encountered significant performance degradation at the customer side, especially when their deployments span multiple data centers over a Wide Area Network (WAN). Our investigation identifies the bottleneck in the performance of the Distributed Hash Join (Dist-HJ) algorithm, which is contingent upon a crucial balance between communication overhead and computational load. This balance is severely disrupted when processing skewed data from real-world customer workloads, leading to the observed performance decline. To tackle this challenge, we introduce Bala-Join, an adaptive solution to balance the computation and network load in Dist-HJ execution. Our approach consists of the Balanced Partition and Partial Replication (BPPR) algorithm and a distributed online skewed join key detector. The former achieves balanced redistribution of skewed data through a multicast mechanism to improve computational performance and reduce network overhead. The latter provides real-time skewed join key information tailored to BPPR. Furthermore, an Active-Signaling and Asynchronous-Pulling (ASAP) mechanism is incorporated to enable efficient, real-time synchronization between the detector and the redistribution process with minimal overhead. Empirical study shows that Bala-Join outperforms the popular Dist-HJ solutions, increasing throughput by 25%-61%.

In practical applications, tower cranes often operate in outdoor environments and are subject to a variety of disturbances. Additionally, the control performance is significantly affected by uncertain dynamics, such as frictional forces and payload mass. To address these challenges, this paper presents a novel adaptive fuzzy control mechanism for nonlinear 5-degree of freedom (5-DOF) tower cranes with variable cable length. The proposed fuzzy adaptive mechanism leverages fuzzy logic to compensate for dynamic uncertainties and external disturbances thereby enhancing the overall robustness of the controlled system. Moreover, to ensure safe operation and avoid collisions, auxiliary constraint terms are introduced to restrict the actuated state variables within predefined bounds throughout the transportation process. In particular, an adaptive law is developed to accurately estimate the payload mass， which is critical for precise cable positioning. To the best of our knowledge, this work proposes a novel adaptive fuzzy closed-loop control framework for 5-DOF tower crane systems with varying cable lengths. The framework systematically integrates online payload mass estimation, state variable constraints, and fuzzy adaptive compensation to improve system stability and robustness. Based on the Lyapunov technique and Barbalat’s lemma, the system stability is theoretically analyzed and proven. Finally, the effectiveness of the proposed controller is validated by hardware experiments.

Antagonistic pneumatic muscle (PM)-actuated wrist robots have great potential in rehabilitation and industrial applications. The antagonistic connection of PMs, which mimics the agonist-antagonist muscle pairs in human joints, provides substantial advantages such as improved joint stability and a better balance of torque disturbances. However, PM-actuated robots exhibit complex nonlinearities, such as hysteresis, creep, input delay, and time-varying parameters, while also confronting challenges such as external disturbances and coupling effects. In this paper, a switching non-singular terminal sliding mode control (NTSMC) method with a double-loop fuzzy neural network (DLFNN) is developed. This method enables the antagonistic PM-actuated wrist robots to achieve fast and precise tracking performance. Specifically, the lumped disturbances are estimated online using the DLFNN, which can adaptively adjust the weight of the inner and outer layers, achieving accurate approximation and robustness. Based on the estimated value of disturbances, a switching NTSMC is implemented to ensure that tracking errors converge to the origin within the fixed time. Switching functions guarantee fast convergence when the sliding surface errors are large. Meanwhile, switching functions ensure non-singularity as the sliding surface errors converge to the origin. Furthermore, joint angles and angular velocities are limited within the specific ranges by designing exponential constraint terms as time-varying proportional-differential gains, rather than traditional barrier functions that may induce excessive control inputs. Both detailed stability analysis and experimental validation demonstrate the effectiveness and adaptability of the proposed method.

Modern database systems rely on workload forecasting to guide performance tuning, resource allocation, and query optimization. However, evolving query patterns and resource demands make accurate workload forecasting increasingly challenging. Most existing methods separately model query and resource behaviors using simple statistics or single-scale forecasting models, thereby ignoring their inherent correlations and leading to inaccurate modeling of real-world workload behaviors. In practice, real-world workloads show that query and resource behaviors are often interdependent and exhibit temporal dynamics. However, most forecasting methods overlook the inherent correlations between query and resource, as well as the temporal patterns across multi-scale resolutions in real workloads. To address these issues, we propose QuadraFormer, an end-to-end forecasting framework that jointly predicts query and resource usage. First, we introduce a unified workload representation that integrates query and resource information into a structured input, enabling joint modeling of workload across variates. Second, we develop an adaptive multi-scale routing mechanism that selects appropriate temporal resolutions to capture periodic patterns and multi-scale variability. Third, we design a novel Quadra-attention mechanism that models both intra- and inter-patch temporal dependencies as well as intra- and inter-dimensional dependencies. Extensive experiments on real-world workloads show that QuadraFormer consistently outperforms competitive baselines, achieving the highest F1 score of 98.38% and accuracy of 86.17%, with F1 improvements of up to 10.28% over the best-performing baseline, while reducing training time by up to 2.16×.

Passenger demand forecasting helps optimize vehicle scheduling, thereby improving urban efficiency. Recently, attention-based methods have been used to adequately capture the dynamic nature of spatio-temporal data. However, existing methods that rely on heuristic masking strategies cannot fully adapt to the complex spatio-temporal correlations, hindering the model from focusing on the right context. These works also overlook the high-level correlations that exist in the real world. Effectively integrating these high-level correlations with the original correlations is crucial. To fill this gap, we propose the Aggregation Differential Transformer (ADFormer), which offers new insights to demand forecasting promotion. Specifically, we utilize Differential Attention to capture the original spatial correlations and achieve attention denoising. Meanwhile, we design distinct aggregation strategies based on the nature of space and time. Then, the original correlations are unified with the high-level correlations, enabling the model to capture holistic spatio-temporal relations. Experiments conducted on taxi and bike datasets confirm the effectiveness and efficiency of our model, demonstrating its practical value. The code is available at https://github.com/decisionintelligence/ADFormer.

Listening head generation aims to synthesize realistic and responsive non-verbal listener head motions that respond to speakers in conversational scenarios. Existing methods typically rely on fixed audio-visual input modalities and predefined emotion labels, limiting their adaptability and expressiveness in real-world scenarios. In this paper, we propose a novel real-time framework, REA-Listener, to generate high-fidelity listening head videos with flexible modality adaptation and dynamic emotion modeling. Specifically, we first propose a Modality-Adaptive Mixture of Experts (MA-MoE) module to encode arbitrary combinations of speaker audio and visual signals into a unified embedding space, ensuring robustness under partial modality conditions. To further enhance the temporal consistency of listener emotion, we present a lightweight emotional head dynamics generator with a multi-modal emotion predictor, which infers listener emotions dynamically from speaker context alongside head motion coefficient prediction. Finally, we employ a 3D-aware renderer based on 3D Gaussian Splatting to produce high-quality listener head videos in real time. With these components, our approach achieves efficient head motion generation at 30fps on a single NVIDIA RTX 3090 GPU, supporting real-time interaction. Extensive evaluations and applications demonstrate that our method outperforms state-of-the-art methods in listening head generation.

Time series anomaly detection plays a crucial role in a wide range of real-world applications. Given that time series data can exhibit different patterns at different sampling granularities, multi-scale modeling has proven beneficial for uncovering latent anomaly patterns that may not be apparent at a single scale. However, existing methods often model multi-scale information independently or rely on simple feature fusion strategies, neglecting the dynamic changes in cross-scale associations that occur during anomalies. Moreover, most approaches perform multi-scale modeling based on fixed sliding windows, which limits their ability to capture comprehensive contextual information. In this work, we propose CrossAD, a novel framework for time series Anomaly Detection that takes Cross-scale associations and Cross-window modeling into account. We propose a cross-scale reconstruction that reconstructs fine-grained series from coarser series, explicitly capturing cross-scale associations. Furthermore, we design a query library and incorporate global multi-scale context to overcome the limitations imposed by fixed window sizes. Extensive experiments conducted on seven real-world datasets using nine evaluation metrics validate the effectiveness of CrossAD, demonstrating state-of-the-art performance in anomaly detection.

In the emerging Industrial Internet of Things (IIoT) applications, federated learning (FL) enables model training without the need to transmit raw data directly. Nevertheless, transmitting model parameters could still reveal private information. To further protect local model parameters, differential privacy combined with FL (DPFL) has been introduced. Nonetheless, adding noise in DPFL can severely impact model performance, especially in non-independent and identically distributed (non-iid) data scenarios typical of IIoT environments. It is necessary to carefully balance privacy preservation and utility. In this article, we propose a robust DPFL scheme leveraging local model partition and sparsification (namely, FedMPS) for heterogeneous IIoT scenarios. The local model is divided into a shared part, which is sparsified before adding noise to mitigate its impact, and a private part that remains on the client. We provide a theoretical analysis of the privacy guarantees. Extensive experiments on common datasets, including Fashion-MNIST, CIFAR-10, and CIFAR-100, demonstrate that the proposed approach achieves a better privacy-utility tradeoff, with a 10%–20% improvement compared to baseline methods, and performs well especially in non-iid scenarios.

In the domain of Industrial Internet of Things (IIoT) applications, methodologies such as meta computing are essential for ensuring timely and efficient data acquisition within intricate environments characterized by multiple edge devices and isolated sensor networks. The Age of Information (AoI), a key metric for evaluating data timeliness and relevance, has emerged as a focal point for improving decision-making and system responsiveness. However, managing these systems in resourcelimited edge environments is challenging, as traditional methods struggle to link sink node selection with data collection path planning, leading to inefficiency and poor AoI performance. This paper develops a collaborative optimization framework based on meta computing, which dynamically coordinates distributed computational resources as a unified virtual system. It combines self-organizing mapping (SOM) with reinforcement learning (RL) to optimize sink node selection and data collection paths using AoI metrics. Experimental results show the method significantly reduces AoI, improves data freshness, and optimizes energy use, highlighting meta computings potential for scalable, efficient IIoT solutions.

The control problem of underactuated mechatronic systems is one of the key representatives of complex nonlinear dynamical systems. Starting from the dynamical structures of underactuated systems is usually one of the most direct and effective ways to design controllers. However, controllers developed for specific dynamic models are often difficult to be directly generalized to other underactuated systems. Moreover, due to the lack of control inputs, designing robust controllers for unmatched disturbances (acting on unactuated states) remains a challenging problem. Therefore, based on the Euler-Lagrange dynamics of multi-input-multi-output (MIMO) underactuated systems, this paper gives four coordinate transformations according to different configurations of the inertia matrix, whichextendsthe Olfati transformation to some extent and finallyunifiesunderactuated systems into normal forms. A sliding manifold and an adaptive neural network sliding mode controller are developed with the derived normal forms, which improves transient performance andavoidsthe chattering problem in traditional sliding mode controllers by combining an estimation error-driven adaptive law, a high-order sliding-mode differentiator, and the super-twisting algorithm. More importantly, the stability is guaranteed by Lyapunov techniques even in the presence ofbothpersistent matched disturbances and asymptotically vanishing unmatched disturbances. Furthermore, the proposed control strategy is applied to overhead cranes and tower cranes, whose superior control performance is verified by hardware experiments.

With the development of intelligent transportation systems, vehicles are exposed to a complex network environment. As the mainstream in-vehicle network (IVN), the controller area network (CAN) has many potential security hazards. Existing deep learning-based intrusion detection methods have security performance advantages, however, they consume too much resources and are therefore not suitable to be directly implemented into the IVN. In this paper, we explore computational resource allocation schemes in the IVNs and propose the LiPar, which is a parallel neural network structure using lightweight multi-dimensional spatial and temporal feature fusion learning to perform intrusion detection tasks in the resource-constrained in-vehicle environment. In particular, LiPar adaptively allocates task loads to in-vehicle computing devices, such as multiple electronic control units, domain controllers, and computing gateways by evaluating whether a computing device is suitable to undertake the branch computing tasks according to its real-time resource occupancy. Experiment results show that LiPar achieves better detection performance, running efficiency, and optimized lightweight model size over existing methods, and can be well adapted to the resource-constrained in-vehicle environment and practically protect the in-vehicle CAN bus security. Code is available athttps://github.com/wangkai-tech23/LiPar

This paper proposes an efficient optimization algorithm to dynamically control the Evolved Packet System (EPS) bearer rates to transport various services between the UE and Evolved Packet Core (EPC) in a LTE Femtocell network scenario. The algorithm is focused on improving the accumulated QoE in the networks and takes joint consideration of limited radio and transport resource so as to leverage the resource management in LTE Radio Access Network (RAN) and transport network. In case of link congestion applications running on TCP tend to share the capacity equally. However, different traffic types have different QoE behaviors and hence sharing the resource equally will lead to a non-optimal aggregated QoE. The proposed algorithm will solve this problem by considering the QoE of individual application flows. We formulate the problem as a convex optimization problem, which maximizes the aggregated QoE, represented by Mean Opinion Score (MOS) value, of all users using different applications, and then solve it using Lagrangian relaxation method. The performance of our algorithms is investigated and evaluated by simulations. The simulations show that our proposed QoE-driven rate shaping algorithm results in a significantly better aggregated QoE compared to the legacy scheme that with fixed rate shaping, especially in heavily congested scenarios. Moreover, a discussion on how often to adjust the shaping rates is given based on the complexity and performance investigations.

In this paper, we study the coordinated QoE-based resource allocation problem for downlink direction in a LTE femtocell cluster, considering the transport network limitation. We first formulate the problem as a convex optimization problem, which maximizes the aggregated QoE of all users and solve it using Lagrangian relaxation method. This approach provides an ideal resource allocation solution but it is not practical, because the allocated bandwidth in the model is assumed to be a real value, while in practice, only an integer amount of Physical Resource Blocks (PRBs) can be assigned to a user. However, it is considered as a benchmark for our new developed algorithms. Afterwards, we propose efficient heuristics, which can be applied in real LTE systems due to their low complexity. The performance of our algorithms is investigated and evaluated by simulations. The simulations show that the coordinated resource allocation scheme considering transport limitation results in a better average QoE compared the scheme that does not take into account the transport network capacity. Furthermore, our proposed heuristics are very close to the theoretical optimal bound.

In this paper, we study the QoE-based resource allocation problem for downlink direction in LTE femtocells, considering the transport network limitation. We first formulate the problem as a convex optimization problem, which maximizes the aggregated QoE of all users and solve it using Lagrangian relaxation method. This approach provides an ideal resource allocation solution but it is not practical, because the allocated bandwidth in the model is assumed to be a real value, while in practice, only an integer amount of Physical Resource Blocks (PRBs) can be assigned to a user. However, it is considered as a benchmark for our new developed algorithm. Then we propose an efficient heuristic, which can be applied in real LTE systems. The heuristic is mathematically proven to provide an optimal solution in case there is no limitation in the transport network. In case of transport network limitation, the simulation results show that it gives almost identical performance (ca. 1%) as the optimal solution in the investigated scenario.

In this paper, we propose a new algorithm to solve the downlink resource allocation problem in LTE networks taking the channel conditions into account. We first formulate the problem as a convex optimization problem, which maximizes the aggregated utility of all users. Different from other papers of the same topic, we construct the utility function based on the Exponential Moving Average (EMA) rate instead of the instantaneous data rate. The advantage of this approach is that it guarantees the users with very bad channel conditions still to be scheduled, even if the number of Physical Resource Blocks (PRBs) is smaller than the number of users. The problem is solved optimally with the Lagrangian decomposition method. Extensive simulations have been carried out to compare our approach to the instantaneous data rate approach and to analyze the influence of the system parameters on the behavior of the algorithm.

Mobile Network Virtualization (NV) is an emerging technique which has drawn increasingly research attention. Network Virtualization enables multiple network operators to share a common infrastructure (including core network, transport network and access network) so as to reduce the investment capital while improving the overall performance at the same time. This is achieved by exploring the multiplexing gain. Similarly, Load Balancing (LB) is a well-known mechanism used in mobile networks to offload excessive traffic from high-load cells (hot spots) to low-load ones within one network operator. This paper aims at investigating the potential gain of applying NV in LTE (Long Term Evolution) networks and compares it with the LB scheme gain. In this paper, we propose an LTE virtualization framework (that enables spectrum sharing) and a dynamic load balancing scheme for multi-eNB and multi-VO (Virtual Operator) systems. We compare the performance gain of both schemes for different applications, e.g. VoIP, video, HTTP and FTP. We also investigate the parameterization of both schemes, e.g. sharing intervals, LB intervals and safety margins, in order to find the optimal parameter settings. The presented results show that the LTE networks can benefit from both NV and LB techniques.

Due to the coexistence of different generations of mobile radio technologies, more and more mobile network operators are keen on deploying multiple radio technologies on a single cell site and carry their traffic over a common transport network. This paper studies such a collocated multiradio system: collocated LTE and HSDPA. Firstly, we investigate the potential gain by using a shared transport network to carry different radio technologies. Furthermore, this paper proposes analytical models to dimension the shared transport network for elastic traffic (TCP-based traffic) from the different radio technologies, by taking traffic characteristics, QoS requirements of each radio technology and the deployed transport QoS function and packet scheduling into considerations. The proposal for modeling the shared transport network is to apply the M/G/R-PS model per QoS class while taking the potential multiplexing gain of bandwidth sharing among the different QoS classes into account. For validating the analytical models, a system simulation model is developed. The analytical results derived from the proposed dimensioning models are validated by comparing with the simulation results for various traffic and network scenarios. Presented results demonstrate that proposed analytical models are well suitable for dimensioning of the shared transport network. With the help of the proposed analytical models, we analyze the potential sharing gain of such a shared transport system under different traffic loads and mix of different radio technologies.