In order to address the issues of real-time performance and the low dependency between feature channels in fabric defect detection networks, this paper proposes the ESE-YOLOv5 network based on YOLOv5. Firstly, to addr...
详细信息
As an emerging power source for new energy vehicles, the proton exchange membrane fuel cell-lithium battery hybrid power system still faces challenges such as difficulty in remaining life prediction and unreasonable e...
详细信息
As an emerging power source for new energy vehicles, the proton exchange membrane fuel cell-lithium battery hybrid power system still faces challenges such as difficulty in remaining life prediction and unreasonable energy allocation management. This study investigates how electrochemical surface area degradation and carbon corrosion in the catalyst layer affect the proton exchange membrane fuel cell output power. An integrated assessment system combining electrochemical surface area degradation and carbon corrosion, which could effectively evaluate degradation in the proton exchange membrane fuel cell degradation is proposed. To assess the lifespan of proton exchange membrane fuel cells, the electrochemical surface area degradation and carbon corrosion are utilized as key indicators, which in turn lead to the development of a semi-empirical model. To solve the problem of energy allocation management, this paper proposes a predictive-feedback optimization method that combines data-driven and model-driven approaches. The method includes a hybrid prediction model of a long short-term memory-convolutional neural network, an improved rapid optimization strategy of grey wolf optimizer-particle swarm optimization, and an attention-enhanced fuzzy neural network controller. The inference capability of the convolutional neural network is utilized to correct the larger prediction error of long short-term memory during short-term drastic changes, realizing the maximization of energy utilization efficiency, minimal hydrogen consumption, minimization of operation cost, and optimization objectives for in-formation containing a wide time horizon. Tests of experiments show that the implementation of this control scheme effectively postponed the deterioration of the proton exchange membrane fuel cell-lithium battery hybrid power system, resulting in an extended lifespan.
Proton exchange membrane fuel cells (PEMFCs) will greatly shorten their lifespan due to platinum (Pt) catalyst degradation during operation. This paper proposes an optimization-based energy management method consideri...
详细信息
Proton exchange membrane fuel cells (PEMFCs) will greatly shorten their lifespan due to platinum (Pt) catalyst degradation during operation. This paper proposes an optimization-based energy management method considering Pt degradation, which is an improvement over the traditional strategy that only focuses on fuel optimization to consider both the minimal fuel and the minimum fuel cell life decay. Firstly, a one-dimensional (1D) Pt degradation model is established to comprehend how various voltage situations affect Pt deterioration. Then, various strategies to suppress Pt degradation are designed using Pontryagin's minimum principle (PMP) optimization algorithm in light of the influence analysis results, and the effects of the PMP algorithm under different strategies are tested on the hardware-in-the-loop (HIL) simulation platform. The results demonstrate that the performance of the PMP algorithm in real-time strategy is extremely near to the global optimal solution generated by the offline dynamic programming (DP) algorithm. After adding the tendency to limit high potential and voltage variation in the PMP algorithm, hydrogen consumption increases by only 2%. In comparison, the stack's degradation is decreased by nearly 50%, considerably extending the stack's service life and reducing the system's comprehensive use cost.
Motor imagery (MI) based brain-computer interfaces (BCIs) enable the direct control of external devices through the imagined movements of various body parts. Unlike previous systems that used fixed-length EEG trials f...
详细信息
Motor imagery (MI) based brain-computer interfaces (BCIs) enable the direct control of external devices through the imagined movements of various body parts. Unlike previous systems that used fixed-length EEG trials for MI decoding, asynchronous BCIs aim to detect the user's MI without explicit triggers. They are challenging to implement, because the algorithm needs to first distinguish between resting-states and MI trials, and then classify the MI trials into the correct task, all without any triggers. This paper proposes a sliding window prescreening and classification (SWPC) approach for MI-based asynchronous BCIs, which consists of two modules: a prescreening module to screen MI trials out of the resting-state, and a classification module for MI classification. Both modules are trained with supervised learning followed by self-supervised learning, which refines the feature extractors. Within-subject and cross-subject asynchronous MI classifications on four different EEG datasets validated the effectiveness of SWPC, i.e., it always achieved the highest average classification accuracy, and outperformed the best state-of-the-art baseline on each dataset by about 2%.
All-in-Focus (AIF) photography is expected to be a commercial selling point for modern smartphones. Standard AIF synthesis requires manual, time-consuming operations such as focal stack compositing, which is unfriendl...
详细信息
All-in-Focus (AIF) photography is expected to be a commercial selling point for modern smartphones. Standard AIF synthesis requires manual, time-consuming operations such as focal stack compositing, which is unfriendly to ordinary people. To achieve point-and-shoot AIF photography with a smartphone, we expect that an AIF photo can be generated from one shot of the scene, instead of from multiple photos captured by the same camera. Benefiting from the multi-camera module in modern smartphones, we introduce a new task of AIF synthesis from main (wide) and ultra-wide cameras. The goal is to recover sharp details from defocused regions in the main-camera photo with the help of the ultra-wide-camera one. The camera setting poses new challenges such as parallax-induced occlusions and inconsistent color between cameras. To overcome the challenges, we introduce a predict-and-refine network to mitigate occlusions and propose dynamic frequency-domain alignment for color correction. To enable effective training and evaluation, we also build an AIF dataset with 2686 unique scenes. Each scene includes two photos captured by the main camera, one photo captured by the ultra-wide camera, and a synthesized AIF photo. Results show that our solution, termed EasyAIF, can produce high-quality AIF photos and outperforms strong baselines quantitatively and qualitatively. For the first time, we demonstrate point-and-shoot AIF photo synthesis successfully from main and ultra-wide cameras.
Depth information plays a pivotal role in numerous computer vision applications, including autonomous driving, 3D reconstruction, and 3D content generation. When deploying depth estimation models in practical applicat...
详细信息
Depth information plays a pivotal role in numerous computer vision applications, including autonomous driving, 3D reconstruction, and 3D content generation. When deploying depth estimation models in practical applications, it is essential to ensure that the models have strong generalization capabilities. However, existing depth estimation methods primarily concentrate on robust single-image depth estimation, leading to the occurrence of flickering artifacts when applied to video inputs. On the other hand, video depth estimation methods either consume excessive computational resources or lack robustness. To address the above issues, we propose ViTA, a video transformer adaptor, to estimate temporally consistent video depth in the wild. In particular, we leverage a pre-trained image transformer (i.e., DPT) and introduce additional temporal embeddings in the transformer blocks. Such designs enable our ViTA to output reliable results given an unconstrained video. Besides, we present a spatio-temporal consistency loss for supervision. The spatial loss computes the per-pixel discrepancy between the prediction and the ground truth in space, while the temporal loss regularizes the inconsistent outputs of the same point in consecutive frames. To find the correspondences between consecutive frames, we design a bi-directional warping strategy based on the forward and backward optical flow. During inference, our ViTA no longer requires optical flow estimation, which enables it to estimate spatially accurate and temporally consistent video depth maps with fine-grained details in real time. We conduct a detailed ablation study to verify the effectiveness of the proposed components. Extensive experiments on the zero-shot cross-dataset evaluation demonstrate that the proposed method is superior to previous methods.
Social interaction, which is intricate and time-varying, has emerged as a pivotal consideration in epidemic spreading. In this paper, we devise a UAU-SIS model for simulating awareness diffusion and epidemic spreading...
详细信息
Social interaction, which is intricate and time-varying, has emerged as a pivotal consideration in epidemic spreading. In this paper, we devise a UAU-SIS model for simulating awareness diffusion and epidemic spreading on temporal multiplex networks. Drawing inspiration from individuals' self-protection behaviors, a novel adaptive update mechanism is developed. To provide a more precise representation of the collective social interactions encompassing individuals, we introduce the higher-order network structure encompassing temporal variability. Building upon the established framework of the microscopic Markov chain approach (MMCA) for static networks, we extend its applicability to the condition of temporal networks and derive the threshold within the coupled dynamics. Our extensive simulations illuminate the dual role of awareness in epidemic mitigation. Beyond solely diminishing infection probabilities through self-protective measures, individual awareness additionally facilitates to change the network structure to separate them from the infected. By elucidating these fundamental characteristics, our research contributes to advancing more effective strategies for epidemic mitigation and containment.
Monitoring devices, as assistance devices for patrols, are being used more frequently in real patrol processes. The deployment of security resources requires full consideration of the coordination between monitoring d...
详细信息
Monitoring devices, as assistance devices for patrols, are being used more frequently in real patrol processes. The deployment of security resources requires full consideration of the coordination between monitoring devices and patrols to ensure a safe environment. The traditional approach to this problem is to deploy monitoring devices and patrols separately based on the evaluation of their respective defensive effect. This paper determines the distribution of monitoring devices and the scheduler of patrols by combining the impact of monitoring devices and patrols. The attack-defense game model is used to distribute monitoring devices and schedule patrols. We determine the optimal patrol strategy using the Markov reward process and game at each moment. The optimal patrol path is composed of the optimal patrol target at each moment, which reduces the number of strategies. The belief state of the partially observable Markov decision process is used to mitigate the issue of incomplete information in the game. The target at the entrance location is shown to be important, and the defender should not place important materials at the entrance.
A novel fault-tolerant system for a multi-motor drive with one-leg, two-leg, and secondary faults is designed in this study. Fault tolerance is realized from pre-fault to post-fault via a topology reconstruction schem...
详细信息
A novel fault-tolerant system for a multi-motor drive with one-leg, two-leg, and secondary faults is designed in this study. Fault tolerance is realized from pre-fault to post-fault via a topology reconstruction scheme and an integrated control law. The scheme develops a reconstructible topology and its reconstruction algorithm. The multi-motor drive can be reconstructed into appropriate tolerant topologies on the basis of the diagnosis results. The control law based on the predictive torque control method is universal under any topology by designing a unified voltage vector model and an optimal switching state decision. The proposed system can cope with more types of faults and effectively switch from a healthy state to a fault-tolerant state under any fault. The effectiveness and performance of the fault-tolerant system are verified and demonstrated by simulation and experiment results.
作者:
Liu, QimingCui, XinruLiu, ZheWang, HeshengDepartment of Automation
Shanghai Jiao Tong University Shanghai China MoE Key Lab of Artificial Intelligence
AI Institute Shanghai Jiao Tong University Shanghai China Department of Automation
Key Laboratory of System Control and Information Processing of Ministry of Education Key Laboratory of Marine Intelligent Equipment and System of Ministry of Education Shanghai Engineering Research Center of Intelligent Control and Management Shanghai Jiao Tong University Shanghai China
Target search in unknown environments places high demands not only on an autonomous vehicle's ability to perceive and interpret target cues, but also on its conscious of collecting these cues by active exploration...
详细信息
Target search in unknown environments places high demands not only on an autonomous vehicle's ability to perceive and interpret target cues, but also on its conscious of collecting these cues by active exploration. While existing navigation methods have successfully built target-driven policies by maintaining memory of explored areas, there has been a lack of focus on facilitating target-aware exploration-the informative frontier information at unexplored yet visible areas is often overlooked. In this paper, we introduce a novel topology-based memory structure, Frontier-enhanced Topological Memory (FTM), and a Hierarchical Topology Encoding and Extraction (HTEE) module, fostering the autonomous vehicle's awareness of both environmental exploration and target approach. Specifically, FTM innovatively incorporates informative ghost nodes on traditional topological map to represent unexplored yet visible regions. We leverage an online-trained implicit scene representation to estimate the positions and generate features of these ghost nodes. The HTEE then employs implicit graph convolutions and attention mechanisms to extract cognitive information from FTM, taking into account the hierarchical memory structure, target cues, and current state. Our design bolsters cognitive navigation decisions. The experiments in the high-fidelity environments, including performance tests, visualizations, and interpretability experiments, validate the effectiveness of our approach in enhancing the vehicle's exploratory behavior. The improved exploration awareness for target cue collection, in turn, enhances the success rate and path efficiency of target search. Furthermore, we demonstrate the adaptability of our algorithm in real-world physical environments. IEEE
暂无评论