In this paper, we investigate an air-to-ground (A2G) wireless network system where multiple uncrewed aerial vehicles (UAVs) provide downlink communication coverage for mobile ground users (GUs). This system accounts for UAVs progressively depleting their energy during coverage provision, ceasing operations when their energy reserves fall below a predefined threshold. We aim to maximize cumulative system throughput over the task period while satisfying the minimum fairness requirement through joint trajectory control and resource allocation (JTCRA) optimization. To meet the fairness requirement, enhancing system robustness is critical; energy-sufficient UAVs must autonomously assist GUs that lose connectivity when their serving UAVs terminate operations. Therefore, we propose a multi-agent deep reinforcement learning (MADRL) framework with a parameter-sharing architecture to solve this problem. As conventional parameter sharing is restricted to homogeneous agents with identical observation-action spaces, we design a dual-agent structure: a trajectory agent (Traj-agent) and a communication agent (Comm-agent) are deployed for each UAV. This separation organizes the heterogeneous tasks of trajectory control and resource allocation into distinct homogeneous agent groups, facilitating effective parameter sharing within each type. Based on this framework, we apply two alternative algorithms: an MAPPO-based JTCRA algorithm and a QMIX-based JTCRA algorithm. Simulation results demonstrate the superiority and effectiveness of our proposed JTCRA algorithms, which maintain service continuity for GUs through intelligent trajectory control, thereby minimizing the adverse impact of coverage gaps.

Optimizing unmanned aerial vehicle (UAV)-assisted wireless networks to serve mobile users (MUs) via beamforming presents significant challenges, mainly due to the dynamic and complex environments. Traditional single-modal data-based modeling methods are often insufficient for capturing the varying environmental characteristics, leading to inaccurate UAV trajectory design and beamforming. To address these issues, we propose a multi-UAV-assisted integrated sensing, communication, and computation (ISCC) framework that processes multi-modal data to enhance environmental awareness and improve communication performance. We then formulate an optimization problem to maximize the average sum rate by jointly optimizing the UAV trajectory and beamforming vectors. Given the non-convex nature of the problem, traditional optimization techniques are inadequate. To this end, we introduce a fine-tuned multi-modal large language model (M2LLM)-driven deep reinforcement learning (DRL)-based joint optimization framework. Specifically, a pre-trained M2LLM is first fine-tuned to predict future MU positions by leveraging historical multi-modal data, including texts, images, and wireless sensing data. The fine-tuned M2LLM is then employed to extract environmental features, where the output of the fine-tuned M2LLM’s last hidden layer is regarded as the environment state vector to eliminate the output uncertainty of the M2LLM. Subsequently, we use a DRL agent to optimize the UAV trajectory and beamforming in a coordinated manner. Extensive simulation results demonstrate that the proposed framework can significantly enhance network performance by enabling environment-aware and adaptive trajectory design and beamforming. The code is available in https://huggingface.co/blYin/MmllmDrlUavTdBf.

In an integrated sensing and communication (ISAC) system, it is often necessary to sense both the state of individual targets and the overall situation of a group target (GT) simultaneously, but the latter is more challenging due to the limited sensing capabilities and resources of a single base station. Owing to the recent rapid development of artificial intelligence (AI) and unmanned aerial vehicle (UAV) technologies, it is feasible to acquire the high-performance situation awareness of the GT by using AI to process sensing data under the cooperation of multiple UAV aerial base stations. However, due to many force majeure factors, such as power depletion, and disruptive actions, some UAVs may be disabled, which affects the situation awareness of the GT. Therefore, it is important to improve the robustness of the situation awareness. To achieve that, we consider a multi-UAV-enabled integrated sensing, communication and computation (ISCC) network, where more than one UAVs complete the group target sensing task (GTST) cooperatively while providing communication services for users. Furthermore, we deploy a pre-trained AI model to process the sensing data for improving the performance of the GTST. To enhance the robustness of the GTST, we apply split learning (SL) to divide the inference task of disabled UAVs into multiple neural network (NN) blocks that are cooperatively inferred by working UAVs. Then, we formulate a GTST completion rate maximization problem in which the trajectory, resource allocation, sensing target scheduling, beamforming, and NN layer split policy are optimized. Due to the non-convexity of the problem, we propose a collaborative multi-agent reinforcement learning scheme. The simulation results show that the proposed scheme can effectively improve the GTST performance while the minimum communication and sensing performance are guaranteed, and its performance is better than that of some benchmark schemes.

Owing to the abundance of onboard energy and wide coverage, fixed-wing unmanned aerial vehicles (FW-UAVs) have better capabilities to serve as aerial base stations, thereby extending communication coverage and improving the performance of ground wireless communication networks. Therefore, the FW-UAV is regarded as one of the essential components of the sixth-generation (6G) communication networks. However, due to its inability to hover, a single FW-UAV may only serve a few mobile users (MUs) at a given time which introduces challenges in ensuring uninterrupted service. Additionally, the limited communication resource further impacts the quality of service (QoS). In order to improve the QoS and guarantee the uninterrupted services of the MUs that are located in a wide range, we consider a multi-FW-UAV communication network to maximize the cumulative throughput by optimizing the trajectory, power control, user association, and subcarrier allocation policy jointly. Since the above problem is non-convex, we first decompose the optimization problem into two subproblems i.e., the trajectory optimization subproblem and the power control, user association, and subcarrier allocation policy optimization subproblem. Then, a multi-agent deep reinforcement learning (MA-DRL)-based joint optimization scheme is proposed to optimize the two subproblems jointly. Simulation results demonstrate that the proposed scheme can maximize the cumulative throughput and gain superior performance compared to the benchmark schemes.

In this work, we investigate an uplink unmanned aerial vehicles (UAVs)-enabled intelligent transportation system to collect data from traveling vehicles on a specific highway road. To ensure the freshness of information delivered from the traveling vehicles to UAV base stations, we use the new age of information (AoI) metric to characterize the information freshness and formulate the AoI minimization problem by optimizing the UAVs’ trajectories and the communication time of vehicles jointly. In order to handle the mixed-integer nonlinear problem, a multi-agent deep reinforcement learning scheme is proposed by applying independent flight direction and time slot action spaces, in which each UAV working as an independent agent adjusts to the dynamic environment quickly based on stored experience. The AoI-related reward function is proposed to select the beneficial action space to guarantee the information freshness. Numerical simulation results show the proposed scheme outperforms the benchmark schemes.

In this work, we investigate a trajectory design problem in uplink unmanned aerial vehicles (UAVs)-enabled data collection system for massive time-sensitive Internet of Things (IoT) services. Although UAV has the advantages of automatic maneuverability and flexible mobility, it is challenging to guarantee the information freshness of collected data under the limited flying energy constraint. Thus we employ Age of Information (AoI) as a new metric to characterize the information freshness and formulate a joint power control and trajectory design optimization problem to minimize average AoI. In order to solve this non-convex problem, we decompose it as a power control subtask and trajectory design subtask, and propose a multi-agent deep reinforcement learning (DRL)-based scheme to solve the subtasks with independent state space, action space and reward function. Simulation results show that the proposed scheme can obtain better performance gain compared to the benchmark scheme and has the superior stability under different settings.

Mobile edge computing (MEC) is a promising technique to meet the demand of computation resources in unmanned aerial vehicle (UAV)-assisted vehicle-to-everything (V2X) networks by offloading computation-intensive tasks to UAV base stations (UBSs). In this paper, we consider an uplink UAV-assisted V2X communication system and characterize the information freshness between UAV and vehicles by applying the new age of information (Aol) metric. To solve the non-convex Aol minimization problem in the high-dimensional action space, a deep Q-network (DQN)-based scheme is proposed to optimize the transmit power and offloading ratio based on the stored experience, in which the action space consists of independent transmit power and offloading ratio. Meanwhile, each UBS selects the beneficial action based on the reward function related to Aol to guarantee the information freshness. The simulation results show that the flight height of UBS and the number of vehicles have a negative impact on Aol. Compared with the benchmark schemes, the proposed scheme can reduce Aol by 40.6%. In addition, the simulation results show that the offloading has a significant effect on Aol compared with the transmit power.

Unmanned aerial vehicle (UAV) as a promising technology in the 6G communication system can collect and transmit information intelligently. However, the existing methods are difficult to design UAV's trajectory to guarantee the information freshness performance. In this paper, the information freshness of a multiple-UAV communication system is modeled based on the age of information (AoI) and the minimization AoI optimization problem subjected to the minimal energy is formulated. In order to solve this nonconvex optimization problem, reinforcement learning (RL)-based scheme is proposed to design the UAVs' trajectory. The proposed scheme constructs the reward function depending on the accumulated AoI to make a fast trajectory decision and reduce the AoI of UAV communication system. The simulation results show that the proposed scheme can improve 21.7% performance gain of information freshness compared to the random scheme and the greedy scheme, and 7.7% performance gain compared to the flying-hover-communication scheme. In addition, the proposed UAV trajectory design scheme has the superior convergence.

In this study, the authors address the problem of parsing fashion images into mid‐level semantic parts including upper‐clothing, lower‐clothing, skin, hair and background. These mid‐level parts provide the regional information of fashion items and have potential value in high‐level parsing process. The key idea of the method is to parse the mid‐level parts by region expanding. Owing to the co‐occurrence of pose skeleton and the proposed parts, the region expanding process starts from the super‐pixels crossed by specific segments of pose skeleton. The super‐pixels are then merged with their neighbours by conditional inference based on their position and perceptual similarity. To avoid the difficulties of training on arbitrary graph structures, conditional random fields (CRFs) are constructed on super‐pixel chains, which are extracted from the generated expanding trees. This is followed by a voting stage to mix up the probabilities estimated by the chain‐CRFs to obtain the final result. Experiments on two datasets show that the new method outperforms related approaches in regional accuracy and has good generalisation capability. Furthermore, the method can be easily employed to improve the performance of high‐level parsing. Its effectiveness has been verified by another group of experiments on two state‐of‐the‐art high‐level parsing approaches.

Texture parsing benefits attribute-based clothing analysis and related applications, such as clothing retrieval and recognition. To deal with the large variations of clothing textures, in this paper, a new method is presented in which refined texture attributes are parsed. Based on the characteristics of clothing textures, refined texture attributes are proposed and parameterized. To estimate the attribute parameters, we exploit the discriminative meanings of sparse codes: the underlying connections between the attribute parameters and each component of sparse codes. The attribute parameters are mapped from the dominating components of sparse codes. Our experiments demonstrate the effectiveness of the proposed method.

In this paper, a novel texture feature GMACM, is presented according to the statistics of gradient angle cooccurrence in color images. Based on three different types of gradients defined in the RGB space, the corresponding GMACMs are introduced. With some well-designed color image classification experiments, it is shown that GMACMs outperform GLCM and Gabor filters significantly in efficiency and accuracy. It could be concluded that GMACM is powerful in classifying and understand color images.

How to make business process flexibly adapted to the changes of organization is the key problem for an enterprise to improve its competitive power. Aimed at providing a solution to this problem, a new flexible workflow model based on extensible organization definition is put forward. The proposed model is composed of organization model and process model. In organization model the new concept of sub-organization has been introduced with the realization mechanism, making the organization model extensible qualified. In process model all sub-organizations and workgroups are allowed to create sub-process independently and refer to sub-process dynamically. Compared with the traditional workflow model with the capability of sub-process definition, the proposed model reduces the effects on business process imposed by the change of enterprise organization structure, gains much more flexibility, and provides management means at all administrative levels