Search

Deep reinforcement learning (DRL) has significantly advanced online path planning for unmanned aerial vehicles (UAVs). Nonetheless, DRL-based methods often encounter reduced performance when dealing with unfamiliar scenarios. This decline is mainly caused by the presence of non-causal and domain-specific elements within visual representations, which negatively impact the learning of policies. Present techniques generally depend on predefined augmentation or regularization methods intended to direct the model toward identifying causal and domain-invariant components, thereby enhancing the model’s ability to generalize. However, these manually crafted approaches are intrinsically constrained in their coverage and do not consider the entire spectrum of possible scenarios, resulting in less effective performance in new environments. Unlike prior studies, this work prioritizes representation learning and presents a novel method for causal representation disentanglement. The approach successfully distinguishes between causal and non-causal elements in visual data. By concentrating on causal aspects during the policy learning phase, the impact of non-causal factors is minimized, thereby improving the generalizability of DRL models. Experimental results demonstrate that our technique achieves reliable navigation and effective collision avoidance in unseen scenarios, surpassing state-of-the-art deep reinforcement learning algorithms.

This study proposes a method for uniformly revolving swarm robots to entrap multiple targets, which is based on a gene regulatory network, an adaptive decision mechanism, and an improved Vicsek-model. Using the gene regulatory network method, the robots can generate entrapping patterns according to the environmental input, including the positions of the targets and obstacles. Next, an adaptive decision mechanism is proposed, allowing each robot to choose the most well-adapted capture point on the pattern, based on its environment. The robots employ an improved Vicsek-model to maneuver to the planned capture point smoothly, without colliding with other robots or obstacles. The proposed decision mechanism, combined with the improved Vicsek-model, can form a uniform entrapment shape and create a revolving effect around targets while entrapping them. This study also enables swarm robots, with an adaptive pattern formation, to entrap multiple targets in complex environments. Swarm robots can be deployed in the military field of unmanned aerial vehicles’ (UAVs) entrapping multiple targets. Simulation experiments demonstrate the feasibility and superiority of the proposed gene regulatory network method.

For swarm robots moving in a harsh or uncharted outdoor environment without GPS guidance and global communication, algorithms that rely on global-based information are infeasible. Typically, traditional gene regulatory networks (GRNs) that achieve superior performance in forming trapping pattern towards targets require accurate global positional information to guide swarm robots. This article presents a gene regulatory network with Self-organized grouping and entrapping method for swarms (SUNDER-GRN) to achieve adequate trapping performance with a large-scale swarm in a confined multitarget environment with access to only local information. A hierarchical self-organized grouping method (HSG) is proposed to structure subswarms in a distributed way. In addition, a modified distributed controller, with a relative coordinate system that is established to relieve the need for global information, is leveraged to facilitate subswarms entrapment toward different targets, thus improving the global multi-target entrapping performance. The results demonstrate the superiority of SUNDER-GRN in the performance of structuring subswarms and entrapping 10 targets with 200 robots in an environment confined by obstacles and with only local information accessible.

Purpose: This study aims to estimate the regional choroidal thickness from color fundus images from convolutional neural networks in different network structures and task learning models. Method: 1276 color fundus photos and their corresponding choroidal thickness values from healthy subjects were obtained from the Topcon DRI Triton optical coherence tomography machine. Initially, ten commonly used convolutional neural networks were deployed to identify the most accurate model, which was subsequently selected for further training. This selected model was then employed in combination with single-, multiple-, and auxiliary-task training models to predict the average and sub-region choroidal thickness in both ETDRS (Early Treatment Diabetic Retinopathy Study) grids and 100-grid subregions. The values of mean absolute error and coefficient of determination (R2) were involved to evaluate the models' performance. Results: Efficientnet-b0 network outperformed other networks with the lowest mean absolute error value (25.61 μm) and highest R2 (0.7817) in average choroidal thickness. Incorporating diopter spherical, anterior chamber depth, and lens thickness as auxiliary tasks improved predicted accuracy (p-value = 6.39×10−44, 2.72×10−38, 1.15×10−36 respectively). For ETDRS regional choroidal thickness estimation, multi-task model achieved better results than single task model (lowest mean absolute error = 31.10 μm vs. 33.20 μm). The multi-task training also can simultaneously predict the choroidal thickness of 100 grids with a minimum mean absolute error of 33.86 μm. Conclusions: Efficientnet-b0, in combination with multi-task and auxiliary task models, achieve high accuracy in estimating average and regional macular choroidal thickness directly from color fundus photographs.

Nowadays, trajectory planning and control technology has been extensively utilized in industry, agriculture, service, construction, infrastructure inspection, military reconnaissance, disaster management, and other areas [...]

This paper mainly discusses the design of wideband and low-height omnidirectional cylindrical dielectric resonator antenna (DRA). It was theoretically found that the TM-mode category of the cylindrical DRA is suitable for the wideband and low-height design using mode-merging method (TM $_{01 \delta }$ and TM $_{02 \delta }$ mode). An improved apertures-fed technique is proposed to excite the TM-mode DRA. The feeding consists of four arc-shaped apertures, and each aperture was fed by the in-phase signal. Based on the E-field distribution, the higher-order TM $_{02 \delta }$ mode of the cylindrical DRA can be excited by adjusting the position of the apertures, and it fuses together with the fundamental TM $_{01 \delta }$ mode to achieve a wideband antenna. Besides, the aperture mode further widen its impedance bandwidth. To demonstrate the idea, a wideband and low-height omnidirectional cylindrical glass DRA was designed for WLAN 5.2-GHz and 5.8-GHz applications. The experimental results show that the proposed low-height DRA ( $h $ = $0.108\lambda _{0}$ ) can achieve an impedance bandwidth of 25.2 % (without aperture mode) and 34 % (with aperture mode), respectively. The bandwidth enhancement of the proposed design is significant, compared to other reported low-height TM-mode DRAs.

Abstract As high amounts of new energy and electric vehicle (EV) charging stations are connected to the distribution network, the voltage deviations are likely to occur, which will further affect the power quality. It is challenging to manage high quality voltage control of a distribution network only relying on the traditional reactive power control mode. If the reactive power regulation potentials of new energy and EVs can be tapped, it will greatly reduce the reactive power optimization pressure on the network. Keeping this in mind, our reasearch first adds EVs to the traditional distribution network model with new forms of energy, and then a multi-objective optimization model, with achieving the lowest line loss, voltage deviation, and the highest static voltage stability margin as its objectives, is constructed. Meanwihile, the corresponding model parameters are set under different climate and equipment conditions. Ultimately, the optimization model under specific scenarios is obtained. Furthermore, considering the supply and demand relationship of the network, an improved technique for order preference by similarity to an ideal solution decision method is proposed, which aims to judge the adaptability of different algorithms to the optimized model, so as to select a most suitable algorithm for the problem. Finally, a comparison is made between the constructed model and a model without new energy. The results reveal that the constructed model can provide a high quality reactive power regulation strategy.

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.

This study addresses dynamic task allocation challenges in coordinated surveillance involving multiple unmanned aerial vehicles (UAVs). A significant concern is the increased UAV flight distance resulting from the assignment of new missions, leading to decreased reconnaissance efficiency. To tackle this issue, we introduce a collaborative multi-target and multi-UAV reconnaissance scheme. Initially, the multitasking constrained multi-objective optimization framework (MTCOM) is employed to optimize task allocation and reconnaissance time in static scenarios. Subsequently, in case of emergency, we iteratively refine the outcomes of static task allocation through an enhanced auction-based distributed algorithm, effectively reducing UAV flight costs in response to new missions, UAV withdrawal, or damage. Simulation results demonstrate the efficacy of our proposed multi-UAV and multi-target cooperative reconnaissance scheme in resolving dynamic task allocation issues. Additionally, our approach achieves a 5.4% reduction in UAV flight distance compared to traditional allocation methods. The main contribution of this paper is to consider a dynamic scenario model involving UAV damage and the emergence of new reconnaissance areas. Then we propose an innovative collaborative multi-target and multi-UAV reconnaissance scheme to address this issue and, finally, conduct experimental simulations to verify the effectiveness of the algorithm.

Quadrotor unmanned aerial vehicles (UAVs) often encounter intricate environmental and dynamic limitations in real-world applications, underscoring the significance of proficient trajectory planning for ensuring both safety and efficiency during flights. To tackle this challenge, we introduce an innovative approach that harmonizes sophisticated environmental insights with the dynamic state of a UAV within a potential field framework. Our proposition entails a quadrotor trajectory planner grounded in a kinodynamic gene regulation network potential field. The pivotal contribution of this study lies in the amalgamation of environmental perceptions and kinodynamic constraints within a newly devised gene regulation network (GRN) potential field. By enhancing the gene regulation network model, the potential field becomes adaptable to the UAV’s dynamic conditions and its surroundings, thereby extending the GRN into a kinodynamic GRN (K-GRN). The trajectory planner excels at charting courses that guide the quadrotor UAV through intricate environments while taking dynamic constraints into account. The amalgamation of environmental insights and kinodynamic constraints within the potential field framework bolsters the adaptability and stability of the generated trajectories. Empirical results substantiate the efficacy of our proposed methodology.

The autonomous navigation of aerial robots in unknown and complex outdoor environments is a challenging problem that typically requires planners to generate collision-free trajectories based on human expert rules for fast navigation. Presently, aerial robots suffer from high latency in acquiring environmental information, which limits the control strategies that the vehicle can implement. In this study, we proposed the SAC_FAE algorithm for high-speed navigation in complex environments using deep reinforcement learning (DRL) policies. Our approach consisted of a soft actor–critic (SAC) algorithm and a focus autoencoder (FAE). Our end-to-end DRL navigation policy enabled a flying robot to efficiently accomplish navigation tasks without prior map information by relying solely on the front-end depth frames and its own pose information. The proposed algorithm outperformed existing trajectory-based optimization approaches at flight speeds exceeding 3 m/s in multiple testing environments, which demonstrates its effectiveness and efficiency.

Speaker recognition methods based on convolutional neural networks (CNN) have been widely used in the security field and smart wearable devices. However, the traditional CNN has many hyperparameters that are difficult to determine, making the model easily fall into local optimum or even fail to converge during the training process. Intelligent algorithms such as particle swarm optimization and genetic algorithms are used to solve the above problems. However, these algorithms perform poorly compared to the current emerging meta-heuristic algorithms. In this study, the dung beetle optimized convolution neural network (DBO-CNN) is proposed to identify the speakers for the first time, which is helpful in finding suitable hyperparameters for training. By testing the dataset of 50 people, it was demonstrated that the accuracy of the model was significantly improved by using this approach. Compared with the traditional CNN and CNN optimized by other intelligent algorithms, the average accuracy of DBO-CNN has increased by 1.22~4.39% and reached 97.93%.

This paper proposes a multi-radial-distance event classification method based on deep learning. To the best of our knowledge, this is the first time that the $\Phi $ -OTDR can tell how far the target event from the sensing fiber is through deep learning approach. The temporal-spatial data matrix collected by the system is filtered by three different band-pass filters to form RGB images as the input of the Inception_V3 network trained by ImageNet dataset. The passband of three band-pass filters is selected by searching the maximum Euclidean distance in the frequency domain. Three kinds of filters with different frequency bands enhance the effective features of data samples in advance. The simulated annealing (SA) algorithm is applied to search the maximum Euclidean distance. Field experiment includes five kinds of events with four different radial distances, where there are 17 subclasses in total, has been carried out. The classification results show that the classification accuracy reaches 86% and the method can tell both the event type and radial distance.

A dynamic phase demodulation algorithm for direct detection Φ-OTDR is proposed and experimentally demonstrated. It offers a way for single channel direct detection Φ-OTDR, which is the simplest and most cost-effective structure, to realize phase demodulation. Through making use of the redundant information between the sensing series from the vibration impacted region, the local extreme points(LEP) of the sensing series can be divided into two types. The second type LEP offers light phase quadrant information, and then the phase can be further reconstructed through anti-trigonometric function. Single frequency and amplitude-modulated waveforms have been correctly demodulated. The demodulated phase amplitude agrees well with theoretical value, which shows a potential application in strain measurement.

Aiming at the shortcomings of the existing control law based on the global information, this article studies the coverage problem of a given region in the plane using a team of USVs. The coverage goal, which is to cover a given search domain using multiple mobile sensors so that each point is surveyed until a certain preset level is achieved, is formulated in a mathematically precise problem statement. The adaptive control law is presented which enables multi-USV to navigate in a complex environment in the presence of unknown obstacles and guarantees that a fully connected multi-USV system attains the coverage goal. In particular, the dangerous area was divided into two different parts in order to enhance the searching efficiency. Finally, simulation results are presented to validate the feasibility and efficiency of the proposed approach.

Inspired by the behaviour of animal populations in nature, we propose a novel exploration algorithm based on Lévy flight (LF) and artificial potential field (APF). The agent is extended to the swarm level using the APF method through the LF search environment. Virtual leaders generate moving steps to explore the environment through the LF mechanism. To achieve collision-free movement in an unknown constrained environment, a swarm-following mechanism is established, which requires the agents to follow the virtual leader to carry out the LF. The proposed method, combining the advantages of LF and APF which achieve the effect of flocking in an exploration environment, does not rely on complex sensors for environment labelling, memorising, or huge computing power. Agents simply perform elegant and efficient search behaviours as natural creatures adapt to the environment and change formations. The method is especially suitable for the camouflaged flocking exploration environment of bionic robots such as flapping drones. Simulation experiments and real-world experiments on E-puck2 robots were conducted to evaluate the effectiveness of the proposed LF-APF algorithm.

Extensive research has been paid for event detection in the past decades. However, the timeliness requirement of real-world applications cannot be satisfied by these approaches. Early event detector is thus proposed recently to deal with this issue. Early detection aims to recognize the target as early as possible, i.e., it can detect partial events and create a monotonous function to rank them. Although important and practical, few studies have been given for early detection due to its complexity. Max-margin Early Event Detector (MMED) is a well-known approach, which achieves satisfied performance in identifying partial events. However, the MMED works in an offline manner and may fail in this era of streaming sequence. In addition, the large memory consumption and high retraining time cost of the MMED are hard to be satisfied in general platform conditions. In this paper, we introduce an online learning technique with max-margin to early event detection. The proposed model could be adapted to the changing data distribution of the streaming sequences. No historical data need to be stored. Therefore, both the memory requirement and retraining time cost are decreased significantly. We evaluate the proposed approach on three benchmark datasets with various complexities. The extensive results demonstrate both the effectiveness and efficiency of the proposed framework.

Glaucoma is the second leading cause of blindness worldwide, and vision loss from glaucoma cannot be reversed. To remedy this issue, in this paper, a user-friendly mobile App, named Yanbao, is developed. Yanbao App is expected to help users conveniently share high-quality glaucoma screening service by using the proposed glaucoma screening algorithm based on clinical parameters. To the best of our knowledge, it is the first App specially designed for screening glaucoma. The main advantages of our App are: 1) the App has been developed for smartphone, which can be conveniently used at any place and at any time; 2) experiments on both public fundus database and real clinical data demonstrate that the App has good detection and classification accuracy; and 3) users' feedback seems to be quite promising in terms of real-time testing and user experience. Thus, the App could be suitable for glaucoma screening in practice to improve the efficiency of clinicians, balance medical resources, and provide better tiered medical services. The Yanbao App can be available from the website http://url.cn/57tk9jT.

Phase-sensitive optical time domain reflectometer (Φ-OTDR) based distributed optical fiber sensing system has been widely used in many fields such as long range pipeline pre-warning, perimeter security and structure health monitoring. However, the lack of event recognition ability is always being the bottleneck of Φ-OTDR in field application. An event recognition method based on deep learning is proposed in this paper. This method directly uses the temporal-spatial data matrix from Φ-OTDR as the input of a convolutional neural network (CNN). Only a simple bandpass filtering and a gray scale transformation are needed as the pre-processing, which achieves real-time. Besides, an optimized network structure with small size, high training speed and high classification accuracy is built. Experiment results based on 5644 events samples show that this network can achieve 96.67% classification accuracy in recognition of 5 kinds of events and the retraining time is only 7 min for a new sensing setup.

The paper proposes a method using convolutional neural network to effectively evaluate the discrimination between face and non face patterns, gender classification using facial images and facial expression recognition. The novelty of the method lies in the utilization of the initial trainable convolution kernels coefficients derived from the zernike moments by varying the moment order. The performance of the proposed method was compared with the convolutional neural network architecture that used random kernels as initial training parameters. The multilevel configuration of zernike moments was significant in extracting the shape information suitable for hierarchical feature learning to carry out image analysis and classification. Furthermore the results showed an outstanding performance of zernike moment based kernels in terms of the computation time and classification accuracy.

Acute hypotensive episodes (AHEs) are one of the hemodynamic instabilities with high mortality rate that is frequent among many groups of patients. This study presents a methodology to predict AHE for ICU patients based on big data time series. The experimental data we used is mean arterial pressure (MAP), which is transformed from arterial blood pressure (ABP) data. Then, the Hilbert-Huang transform method was used to calculate patient's MAP time series and some features, which are the bandwidth of the amplitude modulation, the frequency modulation, and the power of intrinsic mode function (IMF), were extracted. Finally, the multiple genetic programming (Multi-GP) is used to build the classification models for detection of AHE. The methodology is applied in the datasets of the 10th PhysioNet and Computers Cardiology Challenge in 2009 and Multiparameter Intelligent Monitoring for Intensive Care (MIMIC-II). We achieve the accuracy of 83.33% in the training set and 91.89% in the testing set of the 2009 challenge's dataset and the 84.13% in the training set and 82.41% in the testing set of the MIMIC-II dataset.

The mitochondrial expression of exogenous antibiotic resistance genes has not been demonstrated successfully to date, which has limited the development of antibiotic resistance genes as selectable markers for mitochondrial site-directed transformation in Chlamydomonas reinhardtii. In this work, the plasmid pBSLPNCB was constructed by inserting the gene ble of Streptoalloteichus hindustanus (Sh ble), encoding a small (14-kilodalton) protective protein into the site between TERMINVREP-Left repeats and the cob gene in a fragment of mitochondrial DNA (mtDNA) of C. reinhardtii. The fusion DNA-construct, which contained TERMINVREP-Left, Sh ble, cob, and partial nd4 sequence, were introduced into the mitochondria of the respiratory deficient dum-1 mutant CC-2654 of C. reinhardtii by biolistic particle delivery system. A large number of transformants were obtained after eight weeks in the dark. Subsequent subculture of the transformants on the selection TAP media containing 3 ìg/mL Zeomycin for 12 months resulted in genetically modified transgenic algae MT-Bs. Sequencing and Southern analyses on the mitochondrial genome of the different MT-B lines revealed that Sh ble gene had been integrated into the mitochondrial genome of C. reinhardtii. Both Western blot, using the anti-BLE monoclonal antibody, and Zeomycin tolerance analysis confirmed the presence of BLE protein in the transgenic algal cells. It indicates that the Sh ble gene can be stably expressed in the mitochondria of C. reinhardtii.

To realize design automation of mechatronic systems, there are two major issues to be dealt with: opentopology generation of mechatronic systems and simulation or analysis of those models. For the first issue, we exploit the strong topology exploration capability of genetic programming to create and evolve structures representing mechatronic systems. With the use of ERCs (ephemeral random constants) in genetic programming, we can evolve the sizing of mechatronic system components together with the system structures simultaneously. The second issue, simulation and analysis of those system models, is made more complex when the systems are mixed-energy-domain systems. We take advantage of bond graphs as a tool for multi- or mixed-domain modeling and simulation of mechatronic systems. Because there are many considerations in mechatronic system design that are not completely captured by a bond graph, it is beneficial to generate multiple solutions, allowing the designer more latitude in choosing a model to implement. The approach in this paper is capable of providing a variety of design choices to the designer for further analysis, comparison and trade-off study. The approach is shown to be efficient and effective and is demonstrated in an example of open-ended real-world mechatronic system design application, a typewriter re-design problem.

To realize design automation of mechatronic systems, there are two major issues to be dealt with: open-topology generation of mechatronic systems and simulation or analysis of those models. For the first issue, we exploit the strong topology exploration capability of genetic programming to create and evolve structures representing mechatronic systems. With the use of ERCs (ephemeral random constants) in genetic programming, we can evolve the sizing of mechatronic system components together with the system structures simultaneously. The second issue, simulation and analysis of those system models, is made more complex when the systems are mixed-energy-domain systems. We take advantage of bond graphs as a tool for multi- or mixed-domain modeling and simulation of mechatronic systems. Because there are many considerations in mechatronic system design that are not completely captured by a bond graph, it is beneficial to generate multiple solutions, allowing the designer more latitude in choosing a model to implement. The approach in this paper is capable of providing a variety of design choices to the designer for further analysis, comparison and trade-off study. The approach is shown to be efficient and effective and is demonstrated in an example of open-ended real-world mechatronic system design application, a typewriter re-design problem.