Most existing consensus control in multi-agent systems ( MAS s) require agents to update their state synchronously, which means that some agents need to wait for all individuals to complete the iteration before starti...
详细信息
Most existing consensus control in multi-agent systems ( MAS s) require agents to update their state synchronously, which means that some agents need to wait for all individuals to complete the iteration before starting the next iteration. To overcome this bottleneck, this paper studied asynchronous consensus problems of second-order MAS s ( SOMAS s) with aperiodic communication. An asynchronous pulsemodulated intermittent control ( APIMC ) with heterogeneous pulse-modulated function and time-varying control period, which can unify impulsive control and sampled-data control, is proposed for the consensus of SOMAS s. A time-varying discrete system is constructed to describe the evolution of the sample values of position and velocity of the SOMAS . Then, by the analysis tools from the stochastic matrix and the properties of the Laplace matrix of graph, some effective conditions are obtained to show the relationship between the convergence of the controlled SOMAS s and the control parameters. Finally, a 300-node SOMAS whose topology is a random geographic network is included to verify the feasibility of the proposed control and the correctness of the theoretical analysis. & COPY;2022 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.
This letter investigates the problem of simultaneous state and unknown input estimation in a nonlinear system within multi-sensor networks. To avoid the linearization errors caused by existing methods, such as statist...
详细信息
This letter investigates the problem of simultaneous state and unknown input estimation in a nonlinear system within multi-sensor networks. To avoid the linearization errors caused by existing methods, such as statistical linear regression techniques and first-order Taylor approximation, a novel optimal distributed nonlinear filter is proposed based on the unscented transformation and unbiased minimum variance criterion. First, the unscented transformation is applied to generate predicted estimates and the covariance matrix. Second, input and state estimation are carried out in an unbiased minimum variance manner using measurements from the nonlinear system. Third, a distributed strategy leveraging average consensus is employed to incorporate local estimates from neighboring sensors. Finally, simulation results confirm that the proposed method achieves substantial improvements in estimation accuracy compared to existing methods.
Recognizing the physical properties of deformable objects poses great challenges to the density and sensitivity of tactile sensors. Monolithic active layer inevitably introduces large electrical crosstalk to high-dens...
详细信息
Recognizing the physical properties of deformable objects poses great challenges to the density and sensitivity of tactile sensors. Monolithic active layer inevitably introduces large electrical crosstalk to high-density sensors, and the traditional trial-and-error method is inefficient in exploring the sensor recipes with optimum sensitivity. In this work, we present the design and implementation of a high-density flexible tactile sensor array. The structured conductive polymer on parallel electrodes was designed to reduce the electrical crosstalk. Meanwhile, an active learning approach was utilized to efficiently explore the relationship between sensor sensitivity and recipes, so as to find the optimum sensor sensitivity. For applications of the sensor array, a tendon-driven gripper with flexible joints was built, where the sensor array was attached to the fingertip. Experiments on recognizing the size and stiffness of deformable objects were conducted to validate the effectiveness of the sensor array. The results indicate that the design paradigm is expected to promote the exploration and applications of tactile sensors.
Standard approaches for video action recognition usually operate on full input videos, which is inefficient due to the widespread spatio-temporal redundancy in videos. The recent progress in masked video modelling, sp...
详细信息
Standard approaches for video action recognition usually operate on full input videos, which is inefficient due to the widespread spatio-temporal redundancy in videos. The recent progress in masked video modelling, specifically VideoMAE, has shown the ability of vanilla Vision Transformers (ViT) to complement spatio-temporal contexts using limited visual content. Inspired by this, we propose Masked Action Recognition (MAR), which reduces redundant computation by discarding a proportion of patches and operating only on a portion of the videos. MAR includes two essential components: cell running masking and bridging classifier. Specifically, to enable the ViT to perceive the details beyond the visible patches, cell running masking is used to preserve the spatio-temporal correlations in videos. This ensures that the patches at the same spatial location can be observed in turn for easy reconstructions. Additionally, we notice that, although the partially observed features can reconstruct semantically explicit invisible patches, they fail to achieve accurate classification. To address this issue, we propose a bridging classifier that can help fill the semantic gap between the ViT encoded features used for reconstruction and the specialized features used for classification. Our proposed MAR can reduce the computational cost of ViT by 53%. Extensive experiments have demonstrated that MAR consistently outperforms existing ViT models by a notable margin. Notably, we found that a ViT-Large model fine-tuned by MAR achieves comparable performance to a ViT-Huge model fine-tuned by standard training methods on both Kinetics-400 and Something-Something v2 datasets. Moreover, the computation overhead of our ViT-Large model is only 14.5% of that of the ViT-Huge model.
Training deep neural network (DNN) with noisy labels is practically challenging since inaccurate labels severely degrade the generalization ability of DNN. Previous efforts tend to handle part or full data in a u...
详细信息
Living organisms switch their intrinsic biological states to survive environmental turbulence, in which temperature changes are prevalent in nature. Most artificial temperature-responsive DNA nanosystems work as switc...
详细信息
Living organisms switch their intrinsic biological states to survive environmental turbulence, in which temperature changes are prevalent in nature. Most artificial temperature-responsive DNA nanosystems work as switch modules that transit between "ON-OFF" states, making it difficult to construct nanosystems with diverse functions. In this study, we present a general strategy to build multimode nanosystems based on a temperature-responsive DNA strand displacement reaction. The temperature-responsive DNA strand displacement was controlled by tuning the sequence of the substrate hairpin strands and the invading strands. The nanosystems were demonstrated as logic gates that performed a set of Boolean logical functions at specific temperatures. In addition, an adaptive logic gate was fabricated that could exhibit different logic functions when placed in different temperatures. Specifically, upon the same input strands, the logic gate worked as an XOR gate at 10 degrees C, an OR gate at 35 degrees C, an AND gate at 46 degrees C, and was reset at 55 degrees C. The design and fabrication of the multifunctional nanosystems would help construct advanced temperature-responsive systems that may be used for temperature-controlled multi-stage drug delivery and thermally-controlled multi-step assembly of nanostructures. A temperature-responsive three-state switching DNA nanosystem that performs holding, reacting, and resetting at three different temperatures.
Acquisition of labeled training samples for affective computing is usually costly and time-consuming, as affects are intrinsically subjective, subtle and uncertain, and hence multiple human assessors are needed to eva...
详细信息
Acquisition of labeled training samples for affective computing is usually costly and time-consuming, as affects are intrinsically subjective, subtle and uncertain, and hence multiple human assessors are needed to evaluate each affective sample. Particularly, for affect estimation in the 3D space of valence, arousal and dominance, each assessor has to perform the evaluations in three dimensions, which makes the labeling problem even more challenging. Many sophisticated machine learning approaches have been proposed to reduce the data labeling requirement in various other domains, but so far few have considered affective computing. This paper proposes two multi-task active learning for regression approaches, which select the most beneficial samples to label, by considering the three affect primitives simultaneously. Experimental results on the VAM corpus demonstrated that our optimal sample selection approaches can result in better estimation performance than random selection and several traditional single-task active learning approaches. Thus, they can help alleviate the data labeling problem in affective computing, i.e., better estimation performance can be obtained from fewer labeling queries.
Transfer learning, which utilizes labeled source domains to facilitate the learning in a target model, is effective in alleviating high intra- and inter-subject variations in electroencephalogram (EEG) based brain-com...
详细信息
Transfer learning, which utilizes labeled source domains to facilitate the learning in a target model, is effective in alleviating high intra- and inter-subject variations in electroencephalogram (EEG) based brain-computer interfaces (BCIs). Existing transfer learning approaches usually use the source subjects' EEG data directly, leading to privacy concerns. This paper considers a decentralized privacy-preserving transfer learning scenario: there are multiple source subjects, whose data and computations are kept local, and only the parameters or predictions of their pre-trained models can be accessed for privacy-protection;then, how to perform effective cross-subject transfer for a new subject with unlabeled EEG trials? We propose an offline unsupervised multi-source decentralized transfer (MSDT) approach, which first generates a pre-trained model from each source subject, and then performs decentralized transfer using the source model parameters (in gray-box settings) or predictions (in black-box settings). Experiments on two datasets from two BCI paradigms, motor imagery and affective BCI, demonstrated that MSDT outperformed several existing approaches, which do not consider privacy-protection at all. In other words, MSDT achieved both high privacy-protection and better classification performance.
The Euclidean distance error of calibration results cannot be calculated during the hand-eye calibration process of a manipulator because the true values of the hand-eye conversion matrix cannot be obtained. In this s...
详细信息
The Euclidean distance error of calibration results cannot be calculated during the hand-eye calibration process of a manipulator because the true values of the hand-eye conversion matrix cannot be obtained. In this study, a new method for error analysis and algorithm optimization is presented. An error analysis of the method is carried out using a priori knowledge that the location of the augmented reality markers is fixed during the calibration process. The coordinates of the AR marker center point are reprojected onto the pixel coordinate system and then compared with the true pixel coordinates of the AR marker center point obtained by corner detection or manual labeling to obtain the Euclidean distance between the two coordinates as the basis for the error analysis. We then fine-tune the results of the hand-eye calibration algorithm to obtain the smallest reprojection error, thereby obtaining higher-precision calibration results. The experimental results show that, compared with the Tsai-Lenz algorithm, the optimized algorithm in this study reduces the average reprojection error by 44.43% and the average visual positioning error by 50.63%. Therefore, the proposed optimization method can significantly improve the accuracy of hand-eye calibration results.
Small Infrared (IR) target detection plays an important role in flight guidance and target early warning. Most existing solutions we call local contrast (LC) based methods utilize the local contrast between the target...
详细信息
Small Infrared (IR) target detection plays an important role in flight guidance and target early warning. Most existing solutions we call local contrast (LC) based methods utilize the local contrast between the targets and the background to detect small infrared targets. This paper designs a characteristic response curve (CRC) to demonstrate characteristics of LC-based methods and proposes a multi-scale target edge diffusion model (M-TED) for small infrared target detection. First, considering the edge diffusion phenomenon on small infrared targets, the transition region is integrated with the target and the background regions to measure local contrast. In addition, global information can be used to suppress the background. Next, a customized multi-scale strategy based on the CRC is applied to generate an appropriate response map. Finally, small targets are detected via segmentation of the response map. Experiments show that the CRC-customized multi-scale strategy can significantly improve LC-based methods' performance and our M-TED performs best on various evaluations.
暂无评论