Please wait a minute...
Most Download
Most Cited
Links
Visited
Total visitors:
Visitors of today:
Now online:

Current Issue

    15 December 2022, Volume 4 Issue 4
    Review Intelligence
    Artificial intelligence and deep learning methods for solving differential equations: the state of the art and prospects
    Jingwei LU, Xiang CHENG, Fei-Yue WANG
    2022, 4(4):  461-476.  doi:10.11959/j.issn.2096-6652.202255
    Asbtract ( 762 )   HTML ( 105)   PDF (6809KB) ( 886 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    With the rapid advancement of fundamental theories and computing capacity, deep learning techniques have made impressive achievements in many fields.Differential equations, as an important tool for describing the physical world, have long been a focus of interest for researchers in various fields.Combining the two methods has gained popularity as a study issue in recent years.Since deep learning can efficiently extract features from large amounts of data and differential equations can reflect objective physical laws, the combination of the two can effectively improve the generalization ability of deep learning and enhance the interpretability of deep learning.Firstly, the problem of solving differential equations by deep learning was briefly introduced.Then, two types of deep learning methods for solving differential equations were introduced: data-driven and physical-informed methods.Furthermore, the applications of relevant deep learning-based solving methods were discussed.Meanwhile, DeDAO (differential equations DAO), a foundation model for artificial intelligence for science, was proposed to address existing challenges.Finally, conclusions of deep learning methods for solving differential equations were presented.

    Surveys and Prospectives
    Mechanism and data knowledge-driven process monitoring method for neutral leaching in zinc hydro-metallurgical
    Hao REN, Bei SUN, Xiaojun LIANG, Chunhua YANG
    2022, 4(4):  477-490.  doi:10.11959/j.issn.2096-6652.202236
    Asbtract ( 151 )   HTML ( 16)   PDF (12831KB) ( 146 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Neutral leaching can be regarded as the key process of dissolving zinc-calcine in zinc hydro-metallurgy to obtain the zinc-electrolyte, and the variation of external environments and disturbances affect the operation states of the neutral leaching process.To this end, a mechanism and data knowledge-driven process monitoring method for neutral leaching in zinc hydro-metallurgical was proposed.This method firstly started from the physical-chemical reaction mechanism and process mechanism of the neutral leaching process, which can be used to excavate the correlation between the mechanical parameters and the monitoring variables to realize the knowledge-driven selection of key monitoring variables.Secondly, the trend change characteristics of the first-order and second-order key variables were combined to realize the data-driven process monitoring.Finally, this proposed method was applied to the monitoring of the practical neutral leaching process.The results show that this method can effectively realize the monitoring of the zinc hydro-metallurgy neutral leaching process, which can be used to improve the process stability of the neutral leaching process.

    Autonomous Underwater Vehicle
    Autonomous Underwater Vehicle
    2022, 4(4):  491-492.  doi:10.11959/j.issn.2096-6652.2022049-1
    Asbtract ( 179 )   HTML ( 54)   PDF (498KB) ( 336 )   Knowledge map   
    References | Related Articles | Metrics
    Special Topic: Autonomous Underwater Vehicle
    UUV mission re-planning based on threat assessment of uncertain events
    Xiang CAO, Changyin SUN
    2022, 4(4):  493-502.  doi:10.11959/j.issn.2096-6652.2022049
    Asbtract ( 168 )   HTML ( 20)   PDF (6390KB) ( 114 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The mission planning of unmanned underwater vehicles (UUV) is directly related to the efficiency of underwater operations.Due to the complex underwater environment and frequent uncertain events, it is sometimes difficult for UUV to complete the mission according to the initial plan.Aiming at the problem of the UUV mission planning, a mission re-planning strategy based on the threat assessment of uncertain events was proposed.Firstly, the UUV performed initial mission planning according to the distribution of mission points.The self-organizing map algorithm was used to plan the time sequence of accessing multi-mission points and the shortest path for the UUV to visit the mission points.Then, the types of uncertain events were determined in the process of executing missions by the UUV and converted into the form of a Bayesian network.Finally, the Bayesian network was used to evaluate the threat degree of uncertain events.If the threat degree was greater than the threshold, the UUV performed mission re-planning.Otherwise, the UUV continued to perform the mission according to the initial plan.The simulation results of mission planning in a variety of uncertain event scenarios showed that the proposed algorithm could ensure the safety of UUV operations and improve the mission completion rate.

    Multi-AUV cooperative localization in adaptive sampling for marine environmental monitoring
    Jiaxin ZHANG, Senlin ZHANG, Meiqin LIU, Shanling DONG, Ronghao ZHENG
    2022, 4(4):  503-512.  doi:10.11959/j.issn.2096-6652.202250
    Asbtract ( 183 )   HTML ( 23)   PDF (7579KB) ( 211 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Efficient and accurate water quality monitoring is of great significance to the development of marine resources, and Special Topic: Autonomous Underwater Vehicle (AUV) has broad application prospects in marine environmental monitoring.There are problems such as low efficiency, poor reliability, insufficient coverage and poor positioning accuracy when a single AUV performs water quality sampling tasks for ocean scalar field estimation.The multi-AUV-based cooperative localization and adaptive sampling system was proposed.Each AUV in the system broadcasted the collected sampling data to its teammates, and based on the data received, it corrected the location of itself based on the extended Kalman filter.With the collected sampling data, the AUV modeled the environmental scalar field with a Gaussian process and used a differential evolution path planner to plan its subsequent sampling path online.Simulation results showed that the proposed method effectively reduced the positioning error of AUVs, and improved the estimation accuracy of the environmental scalar field.

    Quaternion-based single-vector feedback control for fully-actuated dish-shaped AUV
    Yihang XU, Jian LIU, Changyin SUN
    2022, 4(4):  513-521.  doi:10.11959/j.issn.2096-6652.202251
    Asbtract ( 161 )   HTML ( 18)   PDF (6006KB) ( 143 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The single vector quaternion feedback control algorithm for the Special Topic: Autonomous Underwater Vehicle (AUV) was proposed, which used vectors to represent control input and attitude errors.This algorithm solved the attitude dead zone and multi-turn rotation problem in Euler angle attitude.The above algorithm could realize the attitude and position control in a fully-actuated dish-shaped AUV.In order to demonstrate the efficacy of the algorithm mentioned, the physical model of the dish-shaped AUV was built in the simulation and a digital controller was designed based on quaternion feedback.The experimental results showed that the quaternion feedback control algorithm could make the dish-shaped AUV system move from one arbitrary state to another arbitrary state in the state space.The algorithm regarded the body attitude or position as a single closed loop, thus having larger convergence domain and requiring fewer parameters to tuning than the scheme using Euler angles, which made the debugging of dish-shaped AUV control parameters easier and more convenient in engineering application.

    Underwater image enhancement network based on visual Transformer with multiple loss functions fusion
    Xiaofeng CONG, Jie GUI, Jun ZHANG
    2022, 4(4):  522-532.  doi:10.11959/j.issn.2096-6652.202252
    Asbtract ( 253 )   HTML ( 33)   PDF (13748KB) ( 252 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Due to the absorption and scattering of light in water, the images captured by underwater robots suffer from color distortion and reduced contrast.Aiming at alleviating the quality degradation phenomenon of underwater images, an underwater image enhancement network based on vision Transformer that be trained with multiple losses fusion strategy was proposed.The image enhancement network adopted an encoder-decoder architecture, and could be trained in an end-to-end manner.In order to effectively update the parameters of the network for enhancing underwater images, a linear combination of various losses was adopted as the overall optimization objective, including pixel loss, structure loss, edge loss and feature loss.Quantitative experiments were carried out on two large underwater datasets, and the proposed underwater image enhancement network was compared with 7 underwater image enhancement algorithms.The full reference evaluation metrics peak signal-to-noise ratio and structural similarity were calculated in experiment, and the non-referenced metric underwater image quality measure was also computed.The experimental results showed that the proposed underwater image enhancement network could effectively deal with color distortion and contrast reduction.

    Path parameter consensus-based formation and obstacle avoidance control of Special Topic: Autonomous Underwater Vehicles
    Zhen SU, Dianyong LIU, Dazhi SUN, Xiao LIANG
    2022, 4(4):  533-541.  doi:10.11959/j.issn.2096-6652.202253
    Asbtract ( 132 )   HTML ( 13)   PDF (5159KB) ( 98 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    For large-scale and autonomous underwater cooperative work, the formation and obstacle avoidance control of Special Topic: Autonomous Underwater Vehicle (AUV) were researched.Firstly, the formation error model was constructed using the virtual structure, which transformed the formation control guided by paths into path following control and path parameter synchronization.Then, to solve the path following control between vehicles and virtual reference points, following control laws of the individual vehicle were designed using feedback errors of positions and velocities.To solve the path parameter synchronization under knowing partial state information, formation control laws were designed using path parameter consensus.And considering the obstacle constraints, obstacle avoidance control laws based on velocity correction were employed by the aid of artificial potential function.Simulation results showed that AUVs could follow the desired parametric path with the desired formation and avoid obstacles using the proposed method.

    Machine Learning Methods in AI 3.0
    Machine Learning Methods in AI 3.0
    2022, 4(4):  542-543.  doi:10.11959/j.issn.2096-6652.202240-1
    Asbtract ( 262 )   HTML ( 86)   PDF (745KB) ( 477 )   Knowledge map   
    References | Related Articles | Metrics
    Special Column: Machine Learning Methods in AI 3.0
    A survey of image-based few-shot 3D reconstruction
    Hang YU, Yanwei FU, Boyan JIANG, Xiangyang XUE
    2022, 4(4):  544-559.  doi:10.11959/j.issn.2096-6652.202240
    Asbtract ( 261 )   HTML ( 31)   PDF (12821KB) ( 228 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Few-shot 3D reconstruction is considered one of the classic applications of the third generation of artificial intelligence.In the area of computer graphics and computer vision, few-shot 3D reconstruction has attracted the attention of many researchers during the past several decades because of its wide application scenarios and high research value.The area has grown significantly in recent years after the introduction of deep learning methods.The state-of-the-art methods in image-based few-shot 3D reconstruction were reviewed comprehensively and the series of works of our research group were introduced.The various 3D data types were introduced, and their applicability and general processing procedures in 3D reconstruction were discussed.Furthermore, the most widely used datasets were categorized.Finally, some representative experimental results of common 3D reconstructions were presented, and potential future research directions were proposed.

    Study on NeuroSymbolic learning and its applications
    Yinghao CAI, Hua YANG, Xuan AN, Wenshuo WANG, Yidong DU, Jiatao ZHANG, Zhigang WANG
    2022, 4(4):  560-570.  doi:10.11959/j.issn.2096-6652.202234
    Asbtract ( 235 )   HTML ( 29)   PDF (9108KB) ( 209 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The continuous breakthrough of deep learning in perception has promoted the application of AI in various fields.It is found that we can not meet the requirements without improving the intelligence from perception level to higher cognition level.NeuroSymbolic learning can seamlessly integrate neural network methods, that are good at perception tasks, and logical symbolic methods, that are good at reasoning tasks.Therefore, it is one of the best candidates to achieve high-level cognitive intelligence.A practical framework for NeuroSymbolic learning:NSFOL was proposed.Moreover, three typical applications based on NSFOL: robot motion planning, robot task planning and video evaluation for educational experiment were presented.Experiments show that NSFOL can support these three specific applications successfully.Moreover, these implementations have advantages in learn ability, reasonability, interpretability and generalizability.Hope to stimulate more thinking and research to jointly promote research in NeuroSymbolic learning by sharing our preliminary studies in this direction.

    A hybrid physics-data-knowledge driven approach for human-machine hybrid-augmented intelligence-based system management and control
    Jun ZHANG, Peidong XU, Siyuan CHEN, Tianlu GAO, Yuxin DAI, Ke ZHANG, Hang ZHAO, Jiemai GAO, Yuyang BAI, Jinxing LI, Haoran ZHANG, Xiang LI, Jiuxiang CHEN
    2022, 4(4):  571-583.  doi:10.11959/j.issn.2096-6652.202237
    Asbtract ( 289 )   HTML ( 28)   PDF (10594KB) ( 379 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    The core theories, methods and technologies of contemporary system cognition, management, and control have been transferred to big data and artificial intelligence technology, resulting in a gap between the limitations of current artificial intelligence technology and the needs of complex system cognition, management, and control.As a result, a real need has spawned a new form of artificial intelligence: human-machine hybrid-augmented intelligence form, that is, the cooperation of human intelligence and machine intelligence runs through the process of system cognition, management, control, and so on.Human cognition and machine intelligence cognition are mixed together to form enhanced intelligence form.This form is a feasible and important growth mode of artificial intelligence or machine intelligence.A hybrid physics-data-knowledge (PDK) driven approach for human-machine hybrid-augmented intelligence-based system management and control was proposed.The proposed approach was illustrated by the following: trustworthy distributed data, computing, and algorithm, physics-informed deep learning, hybrid deep reinforce learning incorporating system operation rules, causal analysis, and interpretable AI and virtual digital human.In the context of power system dispatch and control, three examples were used for explaining the applications and technical pathways of the proposed PDK approach.

    Papers and Reports
    Research on metro train driverless system based on man-machine hybrid intelligence
    Benzun HUANG, Dewang CHEN, Zhenfeng HE, Xinguo DENG, Marano GIUSEPPECARLO
    2022, 4(4):  584-591.  doi:10.11959/j.issn.2096-6652.202205
    Asbtract ( 199 )   HTML ( 33)   PDF (3698KB) ( 453 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Based on the development status of subway train driving technology at home and abroad, the necessity of subway train intelligent driving development and research was expounded.In view of the poor interpretability of machine learning algorithms used in current unmanned driving, with an introduction of fuzzy system, a metro train unmanned driving system based on man-machine hybrid intelligence was proposed, which realized man-machine hybrid intelligence in two ways.The subway train driverless system combined with cognitive system was explored, which a future-oriented solution for the realization of strong artificial intelligence subway train driverless system in the real sense was provided.

    Research on three frame difference gesture recognition method based on mixed bone features
    Yongqiang ZHANG, Meilin SONG, Tianhu LIU, Menghua MAN
    2022, 4(4):  592-599.  doi:10.11959/j.issn.2096-6652.202211
    Asbtract ( 138 )   HTML ( 41)   PDF (5716KB) ( 155 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    As an excellent algorithm in gesture recognition, three frame difference detection can solve the “double shadow” problem of two frame difference method to a certain extent.But the three frame difference detection will have holes when recognizing gestures, and it cannot adapt to sudden changes in lighting and so on.To solve this problem, the three frame difference gesture recognition method based on mixed bone features was proposed.Firstly, the binary size of the data set was unified, and the background color inter-ference and computation were reduced.Secondly, the three frame difference of mixed bone features was used to detect and track gestures.Finally, the neural network was used for gesture recognition and classification.This method could effectively train hand characteristics, significantly reduce the interference of background on gesture recognition, and improve the recognition efficiency.The experimental results showed that the minimum recognition rate of this method was 93.38% and the maximum was 99.99% in complex background, which could meet the requirement of robustness.This method provides a new idea for gesture recognition in complex image background.

    Depression recognition based on emotional information fused with attentional mechanism
    Yan CHEN, Xueqin LUO, Wei LIANG, Yongfang XIE
    2022, 4(4):  600-609.  doi:10.11959/j.issn.2096-6652.202221
    Asbtract ( 226 )   HTML ( 69)   PDF (6410KB) ( 183 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the current research on using social media data to predict depression ignoring the characteristics of language style and emotional changes over time and lacking of research on the characteristics of emotional state and post metadata, a depression recognition model based on emotional information fused with attentional mechanism was proposed.Firstly, on the basis of the existing research on depression recognition, the text classification convolutional neural network was used to extract the post information and emotional information for each time period of the user in chronological order, and the attention mechanism was introduced to assign different attention weights to the obtained feature matrix to obtain user post information features and user emotional information features.Next, regular matching was used to extract the emotional tendency information, and the weighted emotional feature matrix output by the attention mechanism was spliced to enhance emotional learning expression.Then, metadata features describing social network posts were added, indicators that characterized user preference features were designed, and user language preference features through statistical characterization indicators were extracted.Finally, the three characteristics of user language information, user emotional information, and user language preference information were combined to establish a prediction model for the user’s depression state based on a multi-layer perceptron.The experimental results showed that the accuracy of the model in this paper had increased by 0.051, the recall rate had increased by 0.065, and F1 value had increased by 0.058.

    Lung cell image segmentation method combining Attention U-Net and bottleneck detection
    Hong SHAO, Changsheng ZUO, Ping ZHANG
    2022, 4(4):  610-616.  doi:10.11959/j.issn.2096-6652.202207
    Asbtract ( 98 )   HTML ( 51)   PDF (7249KB) ( 117 )   Knowledge map   
    Figures and Tables | References | Related Articles | Metrics

    Lung pathological images are characterized by fuzzy boundary and overlapping and interweining cells.In order to solve the problem of cell segmentation, a lung cell image segmentation method combining Attention U-Net and bottleneck detection was proposed.Firstly, bilateral filtering and Laplacian sharpening were performed on the collected images to highlight the details of cell edges and increase the contrast between the target and the background while removing the noise.Then the Attention U-Net was trained, and the pathological images were segmented using the trained model to obtain the cell regions.Based on the segmentation results of the model, the discriminant model was established with area, circumference and roundness as screening conditions to distinguish single cell from overlapping cells.The bottleneck detection method was used to determine the separation point in the overlapping region of cells, and the ellipse fitting method was used to modify the boundary, and the final segmentation result was obtained.Experimental results show that this method can segment complex lung cell pathological images (including single cell and overlapping cells) and achieve good segmentation results.