Accurate channel state information (CSI) prediction is essential for 6G massive multiple-input multiple-output (m-MIMO) IoT systems. Deep learning models, such as Transformers, exhibit quadratic computational complexity, resulting in significant efficiency bottlenecks when processing high-dimensional, long sequences channel data in high-mobility scenarios. The Mamba architecture, distinguished by its unique selective state-space model (SSM), presents a promising solution which combines linear computational complexity with robust capabilities for modeling long-range dependencies. Building on this foundation, we propose ChannelMamba, an end-to-end model specifically designed for channel prediction. First, the model employs a dual-domain input module that captures comprehensive channel features by concurrently processing frequency-domain CSI and delay-domain channel impulse response (CIR) data. Sub-sequently, we develop a cross-path parameter-sharing strategy for the Mamba modules to efficiently capture temporal channel dynamics while enhancing model generalization. Furthermore, to address the multi-dimensional dependencies and global context inherent in channel data, we design a bidirectional Mamba module for cross-feature modeling, enhanced with a lightweight attention mechanism. Finally, extensive experimental evaluations across various standard scenarios demonstrate the significant advantages of ChannelMamba over baseline methods in terms of prediction accuracy, robustness, generalization and computational efficiency, achieving new state-of-the-art performance in channel prediction tasks.

Highly flexible Automated Guided Vehicles (AGVs) are interconnected via Industrial Wireless Control Networks (IWCNs) in Multi-access Edge Computing (MEC)-assisted smart factories. The MEC alleviates the lack of computational resources in AGV systems through task offloading. However, IWCNs with limited communication resources struggle to support the highly concurrent offloading of AGVs. In the distributed AGV systems with multi-MEC servers, AGV mobility leads to uneven distribution across MEC server areas, potentially resulting in severe competition for communication resources. Therefore, in this paper, we design a Transferable joint Task Offloading and Multi-Channel Access (T2OMCA) algorithm based on multi-agent deep reinforcement learning. Specifically, AGV observations are modelled as graphs, in which edge relationships are learned through Transformer. This enables AGVs to utilize domain information to collaborate and alleviate concurrent offloading. Moreover, the T2OMCA algorithm converts network input into fixed embeddings to accommodate varying numbers of AGVs. Finally, to encourage exploration in the high-dimensional action space, the T2OMCA algorithm introduces a noisy network and a prioritized experience replay mechanism. Extensive simulations show that the T2OMCA algorithm outperforms existing algorithms in terms of average completion rate, processing delay, and access conflict rate under time-varying AGV topologies.

Accurate trajectory prediction for all agents within complex environments is a crucial step toward realizing autonomous driving navigation. However, this task poses significant challenges due to the uncertainty surrounding the agent's intentions and the intricate road topology. Existing trajectory prediction methods struggle to strike a balance between accuracy and efficiency. To address this challenge, we propose the graph-based trajectory prediction network (DGATP). The model utilizes a two-layer graph representation to capture both the geometric and topological features of the driving environment information and encodes the static and dynamic driving environments hierarchically. An inter-layer network employing an attention mechanism is employed for feature aggregation, leading to improved local-global feature fusion. Furthermore, we introduce a joint prediction framework for all agents in the scenario, which utilizes dynamic weight learning. This adaptive head enhances the model's capacity without increasing its size, thereby maintaining the efficiency of the inference process and leading to accurate and efficient trajectory predictions.

Batteries of electronic quantities detection and state of health have always been the core of the battery management system of electric vehicles, which is capable of estimating SOC accurately and quickly and ensuring the safe operation of electric vehicles. Aiming at the problem of large estimation deviation of SOC and SOH in the whole life cycle of lithium battery, this paper proposes a multi-innovation dual Unscented Kalman Filter based on fractional-order model. Firstly, the fractional-order model of lithium battery is established and the parameters of the model are identified by a genetic algorithm. Secondly, the fractional-order multi-innovation Unscented Kalman Filter is proposed to estimate SOC, and the ohmic resistance and SOH are estimated by Unscented Kalman Filter to improve the SOC estimation accuracy in the whole life cycle. Finally, the proposed algorithm is verified by Urban Dynamometer Driving Schedule(UDDS) dynamic condition data.

Collaborative unmanned aerial vehicle (UAV) networking has the characteristics of flexibility, efficiency, ubiquity, etc., which can enhance wireless network coverage and improve the quality of service to ground users. However, collaborative UAV networking has main challenges such as location deployment and energy optimization. For collaborative networking scenarios, it is necessary to focus on solving key issues such as algorithm convergence and time complexity. In response to the above problems, we propose to leverage hybrid deep reinforcement learning (DRL) with spatial attention mechanism to reduce time complexity according to actual application requirements of location deployment. Firstly, the convolutional neural network is used to improve the ability of discriminating state features. Then, an improved spatial attention mechanism is introduced to apply different weights to state features of different spatial locations, and focuses on the state features that are favorable for UAV deployment in spatial locations. Finally, the offline training state attention model is added to the state input of hybrid deep reinforcement learning for adaptive deployment training. The simulation results show that the training time of the algorithm can be greatly reduced while both mean opinion score (MOS) and energy consumption performance are reduced by about 10%.

Vehicle trajectory prediction technology is of great significance in autonomous driving and intelligent transportation systems. Ego-vehicles can judge the future motion state considering nearby vehicles by predicting their trajectories, which facilitates safe and effective decisions to avoid collisions. It is a challenging task to accurately predict the future trajectories of surrounding vehicles. To solve this problem, we propose a Self-Attention Social Generative Adversarial Networks (SA-SGAN) model to predict trajectories of surrounding vehicles. We use the Self-Attention mechanism to capture the correlation between the features in the vehicle trajectory sequence to effectively solve the problem of missing important information due to a long input sequence, and use training characteristic of Generative Adversarial Networks (GAN) to effectively learn the distribution of real trajectory data and improve prediction accuracy. We evaluate the proposed model through NGSIM dataset, use the trained model to investigate the vehicle trajectory in the next 5s in a three-segment scenario of the US-101 highway, and use the Average Displacement Error (ADE) and Final Displacement Error (FDE) as the evaluation indicators. Compared with baseline methods, the proposed model reduces the evaluation indicators to 4.97 and 8.92 respectively.

Clustering has been extensively used in Vehicular Ad- hoc NETworks (VANETs) for routing optimization and radio resource management, and continues to be considered to facilitate data dissemination in heterogeneous vehicular networks with the ever- increasing data traffic demands. Most of the existing clustering mechanisms in VANETs operate in a distributed mode. However, there is redundant control overhead and transmission decisions, such as cluster maintenance, parameter tuning and forwarding scheduling, which are costly in distributed modes. In this paper, a centralized clustering based hybrid vehicular networking architecture (CC-HVNA) is proposed, in which the collaborative control between IEEE 802.11p and LTE is realized to achieve clustering and to coordinate message delivery. In CC-HVNA, a volatile node state SN is set to reflect ever-changing network topology and to update clusters. Location-based Vehicle to Infrastructure (V2I) communications are utilized to gather regional information so as to perform centralized clusters partition and maintain cluster info table in infrastructures. We leverage a control center to integrate cluster info from the Evolved Node (eNodeB) and Road Side Units (RSUs). Owing to the possession of global cluster info, cluster changes can be detected and targeted data dissemination can be supported according to content-oriented service. The performance evaluation demonstrates that the proposed CC-HVNA clustering scheme can achieve a significant improvement of safety data dissemination.