This paper deals with the problem of clusters flocking for a heterogeneous multi-agentsystem of Unmanned Aerial Vehicles (UAVs) and Unmanned Surface Vessels (USVs). The dynamics models of the UAVs and USVs are establ...
详细信息
ISBN:
(数字)9798350353594
ISBN:
(纸本)9798350353600
This paper deals with the problem of clusters flocking for a heterogeneous multi-agentsystem of Unmanned Aerial Vehicles (UAVs) and Unmanned Surface Vessels (USVs). The dynamics models of the UAVs and USVs are established separately. Due to the different state dimensions and dynamics models of the UAVs and USVs, the two types of agents are formulated with two clusters. Then, the functions of formation errors of the two clusters are constructed. For the two clusters, a new type of prescribed-time formation control laws is designed respectively. Moreover, compared with the conventional formation control that can only ensure that the multi-agentsystem completes ideal formations without strict time constraints theoretically, the present prescribed-time control laws can enable the multi-agentsystem to complete ideal formation strictly at a prescribed time independent of initial formation errors. Finally, the effectiveness of the prescribed-time formation control is verified by simulations.
Achieving joint learning of Salient Object Detection (SOD) and Camouflaged Object Detection (COD) is extremely challenging due to their distinct object characteristics, i.e., saliency and camouflage. The only prelimin...
详细信息
In recent years, rapid progress of deep learning has resulted in object detection algorithms based on deep neural networks nearly completely replacing traditional methods. This shift has led to significant improvement...
In recent years, rapid progress of deep learning has resulted in object detection algorithms based on deep neural networks nearly completely replacing traditional methods. This shift has led to significant improvements in both accuracy and speed of object detection. However, deep neural networks that support decision-making inobject detection are complex black boxes with hidden internal logic and operations. This lack of transparency results in deep learning models being unexplainable, which limits their practical deployment. In this paper, we analyze the current state and future trends of interpretable technology in object detection and propose an effective object detection interpretable algorithm. This algorithm aims to explain the underlying logic of AI decision-making and provide support for the high performance and credibility of object detection algorithms based on deep neural networks through quantitative evaluation indicators.
In non-cooperative scenarios, wideband frequency hopping (FH) communication reconnaissance including FH signal detection, parameter estimation, and network sorting under single-channel reception is challenging. The FH...
In non-cooperative scenarios, wideband frequency hopping (FH) communication reconnaissance including FH signal detection, parameter estimation, and network sorting under single-channel reception is challenging. The FH pattern contains the most information about FH signals. So it is the core of FH parameter estimation. Based on the task analysis, this paper proposes a blind prediction framework combining deep learning (DL) and clustering methods to improve the accuracy and efficiency of FH pattern estimation for mixed signals in the wideband spectrum. The unique advantage of this framework is that it requires no prior signal information and anchors but exploits the inherent time-frequency (TF) properties of asynchronous FH signals. It has a strong generalization ability and can adapt to signals of any shape. Moreover, a new comprehensive evaluation metric, normalized root mean square error including missed detection (NRMSE-MD), suitable for sequence data prediction is proposed to evaluate the level of missed detection and false detection in signal monitoring. Experimental results demonstrate the superiority of the proposed framework and the effectiveness of the proposed evaluation metric.
The upper limb motion intention recognition method based on electroencephalography (EEG) and surface electromyography (sEMG) fusion has achieved significant results in fields such as prosthetic control. However, most ...
The upper limb motion intention recognition method based on electroencephalography (EEG) and surface electromyography (sEMG) fusion has achieved significant results in fields such as prosthetic control. However, most existing fusion methods use manual means to select features, which cannot capture temporal and spatial features at different scales, and ignore the correlation features between the two types of signals. To address these issues, this article proposes a fusion recognition method for upper limb motion intention EEG and sEMG based on Multi-scale Convolution, Polarized Self-Attention, and Cross Intelligence Integration Module. This article extracts multidimensional temporal and spatial features of EEG and sEMG through multi-scale convolution, and introduce polarized self-attention mechanism to filter the extracted multi-scale features. Simultaneously using cross enhancement strategy to extract correlation features between EEG and sEMG. Finally, the features are input into the classification network for recognition. This method was validated on the Jeong database, and the results showed that compared with CNN-LSTM and EEGNet, the recognition accuracy of this method increased by 2.63% and 3.15%, respectively.
Aiming at the problem that the current depth estimation of single image mostly uses the ground public data set, and there is less research on aerial images, this paper uses the collected visible and infrared aerial im...
详细信息
Aiming at the problem that the current depth estimation of single image mostly uses the ground public data set, and there is less research on aerial images, this paper uses the collected visible and infrared aerial image data sets to study the depth estimation. We used FCRN and LapDRN to extract the depth estimation results of visible aerial image and infrared aerial image under global and single object respectively, and compared the results extracted from the two types of networks by qualitative and quantitative methods. The results show that the depth estimation accuracy of visible aerial image is higher than that of infrared image, but the increase rate of infrared image accuracy index is higher than the increase rate of image depth range. The image depth estimation results extracted by LapDRN method are superior to FCRN in accuracy, structure consistency and edge clarity, which indicates that LapDRN is more effective in the estimation of aerial image depth information.
In this paper, the event trigger mechanism-based robust control problem is investigated for cyber-physical systems under denial of service attacks and disturbances. The effects of DoS attacks are regarded as a type of...
详细信息
In this paper, the event trigger mechanism-based robust control problem is investigated for cyber-physical systems under denial of service attacks and disturbances. The effects of DoS attacks are regarded as a type of data packet loss. An event trigger mechanism is introduced into the control scheme design to effectively compensate the packet loss data. The robust attenuation technology is composited into control scheme to reduce the impact of disterbances. Then, by modeling the cyber-physical systems under denial of service(DoS) attack as a class of bounded sequential switching system, and constructing the corresponding Lyapunov function, the stability of the system is verified and analyzed. Finally, simulation experiments are given to verify the correctness of the above theoretical results.
The task allocation for multiple robots is a critical component in the coordination of unmanned clusters. The existing heuristic algorithms are hard to achieve satisfactory results in large-scale problems, and reinfor...
详细信息
ISBN:
(数字)9798350384185
ISBN:
(纸本)9798350384192
The task allocation for multiple robots is a critical component in the coordination of unmanned clusters. The existing heuristic algorithms are hard to achieve satisfactory results in large-scale problems, and reinforcement learning-based methods face challenges in ensuring the rationality of reward design. This paper introduces a model based on multi-head attention and graph neural networks to address the schedule-dependent multi-robot task allocation problem, trained using unsupervised learning techniques. This model can be trained with varying numbers of robots and tasks without necessitating changes to its structure or parameters. In the experiment of this paper, the model is trained under two different conditions, and the performance is evaluated across six different problem scales. Comparing the proposed model against greedy algorithms and genetic algorithms, the results demonstrate that the proposed model has sianiflcant advantages in overall performance.
In the intelligent transportation system, vehicle detection is one of the essential technologies in obstacle avoidance and navigation, however the existing vehicle detection methods cannot meet the actual needs. This ...
详细信息
For the problem of collision volume change when forklift unmanned vehicles work in complex storage environments, a multi-level graph search path planning method is proposed. The first level is the path planning of the...
详细信息
ISBN:
(数字)9798350384185
ISBN:
(纸本)9798350384192
For the problem of collision volume change when forklift unmanned vehicles work in complex storage environments, a multi-level graph search path planning method is proposed. The first level is the path planning of the Unmanned Ground Vehicle (UGV) body, which uses the grid map and the A * algorithm to plan the path of the UGV transferring in the scene; the second level is the joint path planning of the UGV and the object being carried, to avoid overcomplicating this issue, the unmanned vehicle will be limited to pushing the cargo only forwards. The third level is the path planning when the unmanned vehicle carries multiple objects, which decomposes the whole scene into multiple sub-scenes, plans the paths of the unmanned vehicle in the sub-scenes separately, and searches for the optimal combination of sub-scenes. Higher level searches will invoke lower level algorithms and provide a virtual scene and target as input. The experimental results show that this multilevel path planning method is optimal and complete in solving the forklift unmanned vehicle for handling a single object, and can give a feasible solution for handling multiple objects.
暂无评论