Content of Engineering and Applications in our journal

        Published in last 1 year |  In last 2 years |  In last 3 years |  All
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Choice of Coordination Contracts in Live Streaming Supply Chain Considering Streamer Reputation
    PENG Liangjun, LIU Yawei, ZOU Zichen, LIU Mingwu
    Computer Engineering and Applications    2023, 59 (11): 251-262.   DOI: 10.3778/j.issn.1002-8331.2210-0024
    Abstract15)      PDF(pc) (735KB)(3)       Save
    The coordination of the live streaming supply chain is an important guarantee for promoting the development of the digital economy and maintaining the resilience and stability of the supply chain. Aiming at the problem of coordinating-contract selection in the live streaming supply chain, a two- echelon supply chain game model consisting of a manufacturer and a live streaming platform is constructed, and the optimal decision-making and profit differences of the live streaming supply chain under cost sharing and revenue sharing contracts are compared and analyzed, and the impact of the contract on the optimal decision-making and profit of the live streaming supply chain through negotiation and non-negotiation is explored. The study finds that:(1)The live-streaming effort level under the revenue sharing contract is higher than that under the cost sharing contract, but the wholesale price is lower than that under the cost sharing contract, and which retail price is higher under the two contracts depends on a threshold value. (2)Whether it is the profit of supply chain members or the overall profit, the revenue sharing contract is better than the cost sharing contract. (3)The reputation of streamer positively affects the live-streaming efforts and retail prices, and the impact on members profit and overall profit of the live streaming supply chain decreases and then increases. (4)Compared with the non-bargaining situation, in the case of equal bargaining power, the overall profit of the live-streaming supply chain and the profit of manufacturer under both contracts increase, while the profit of the live streaming platform decreases. When the bargaining power is not equal, the overall profit of the live streaming supply chain under both contracts increases with the increase of the bargaining power of the manufacturer.
    Reference | Related Articles | Metrics
    Research on Evacuation Path Planning of Congested Environment with Improved Ant Colony Algorithm
    HUO Feizhou, GAO Shuaiyun, WEI Yunfei, MA Yaping, WU Lijun
    Computer Engineering and Applications    2023, 59 (11): 263-271.   DOI: 10.3778/j.issn.1002-8331.2206-0493
    Abstract15)      PDF(pc) (945KB)(2)       Save
    Aiming at the influence of personnel congestion on evacuation route selection in the process of emergency evacuation, an improved ant colony algorithm evacuation path planning model in congested environment is proposed. Based on the two-dimensional grid environment, trap grids are identified, the diagonal grid environment model is established, and the initial pheromones are differentiated to improve the problem of blind search in the initial stage of the ant colony algorithm. The heuristic function is improved by combining the degree of path congestion and the influence of the end point on the ant path selection, avoiding falling into local optimum and improving the quality of the search path. The pheromone attenuation coefficient is introduced to penalize the path passing through the congested areas, and the suboptimal path obtained by Dijkstra algorithm is combined to improve the update method of pheromone. Through the shortest path optimization operation, the invalid nodes and redundant inflection points of the shortest path are reduced, and the smoothness of the path is improved. The comparative analysis of the simulation experimental results shows that the improved ant colony algorithm can quickly and efficiently plan a smoother and optimal evacuation route with or without congestion.
    Reference | Related Articles | Metrics
    Research on Influence of Government Subsidies on Cross-Border E-Commerce Supply Chain Cooperation Mode
    WANG Xin, YU Lin, ZHANG Shuhua, WANG Xinyu
    Computer Engineering and Applications    2023, 59 (11): 272-284.   DOI: 10.3778/j.issn.1002-8331.2210-0200
    Abstract14)      PDF(pc) (1040KB)(4)       Save
    Against the background of the COVID-19, cross-border e-commerce has further become an important channel for trade market development. The supply chain composed of an export retailer, a cross-border e-commerce platform and foreign consumers is studied. When the government gives or does not give subsidies to the platform, three cooperation modes between the platform and the export retailer are considered, that is, non cooperation mode, cost sharing mode and fixed cost mode. A Stackelberg game model among the export retailer, the platform and the government is established, and the decision-making results under each mode are analyzed and compared. The results show that whether the government provides subsidies for platforms has an important impact on the choice of cooperation modes between platforms and retailers. When the cost coefficient of platform promotion is large, the government should provide subsidies to the platform. When the government provides subsidies to platforms, cross-border e-commerce platforms and export retailers are more inclined to choose the non cooperative mode.
    Reference | Related Articles | Metrics
    Research on Precipitation Estimation Algorithm from Fengyun-4 Satellite Based on Improved U-Net
    HUANG Jie, ZHANG Yonghong, MA Guangyi, ZHU Linglong, TIAN Wei
    Computer Engineering and Applications    2023, 59 (11): 285-293.   DOI: 10.3778/j.issn.1002-8331.2202-0255
    Abstract15)      PDF(pc) (860KB)(2)       Save
    Aiming at the problems of low accuracy and low spatial-temporal resolution of precipitation estimation using satellite images under the condition of severe convective weather, an improved U-Net precipitation estimation algorithm is proposed. Firstly the encoder of the U-Net model is combined with the decoder through the residual module, so that the model parameters can be shared to avoid the disappearance of deep network model gradients. Based on this structure, the spatial pyramid module is introduced for multi-scale feature extraction to retain more image features and strengthen the feature extraction ability of small precipitation cloud information. The attention mechanism module is added to extract important precipitation feature information. The experimental results show that probability of detection, false alarm ration, critical success index of the proposed algorithm are 0.84, 0.48 and 0.59, respectively. The root mean square error and mean absolute error are 1.354?mm/h and 0.432?mm/h respectively. Compared with PERSIANN-CNN and U-Net, the proposed algorithm effectively improves the accuracy of precipitation estimation. Compared with other precipitation products, it also has certain advantages. Therefore, the algorithm can achieve near-real-time results and effectively improve the accuracy of precipitation estimation, which is valuable for the research of precipitation estimation with low tempotal resolution.
    Reference | Related Articles | Metrics
    Double Archive Particle Swarm Optimization Solving Flexible Job-Shop Scheduling Problem
    ZHANG Yujia, SONG Wei
    Computer Engineering and Applications    2023, 59 (11): 294-301.   DOI: 10.3778/j.issn.1002-8331.2202-0298
    Abstract16)      PDF(pc) (588KB)(2)       Save
    This paper proposes a particle swarm optimization based on credibility of solution and double archive to minimize makespan of flexible job-shop scheduling problem(FJSP). Firstly, the elite archive and the optimization archive are constructed to store the individual historical optimal position(Pbest) of the elite particle with better fitness value and the position of the particles that have made rapid progress, respectively. Secondly, it uses the particles in the elite archives to calculate the  credibility of solution, and judge the evolutionary state of the current  population according to the credibility of solution. The particles adjust the strength of learning from the two archives according to the evolutionary state to achieve a balance between convergence and diversity. Further, extensive experiments are carried out on 5 test problems of Kacem and 10 test problems of MK series, and the comparison with other algorithms according to the minimum completion time and average completion time proves the effectiveness of DAPSO in solving the FJSP problem.
    Reference | Related Articles | Metrics
    Taxi Destination Prediction Based on Convolution, Attention and Multilayer Perceptron
    YU Danqing, WU Qunyong, YAO Jiangtao, KUANG Jiaheng
    Computer Engineering and Applications    2023, 59 (11): 302-311.   DOI: 10.3778/j.issn.1002-8331.2203-0168
    Abstract14)      PDF(pc) (805KB)(2)       Save
    Taxi destination prediction based on location services is an important aspect for reasonable planning of urban transportation. However, there exist problems of data sparsity and single feature of trajectory data, which affect the accuracy of destination prediction. To address the problem of data sparsity, the original taxi trajectory data(i. e. taxi speed, direction angle, and time) is combined with trajectory truncation method to determine the model input features, thus overcoming data sparsity. To address the problems of excessive parameters and overfitting caused by a large number of parameters of the multilayer perceptron(MLP), a parameter sharing mechanism based on convolution is proposed to solve parameter redundancy. Furthermore, an attention mechanism is used to allocate more computing resources to more important tasks, focusing on important information to improve the prediction performance of the model. Based on this, a taxi destination prediction method(CCMLP) that combines convolution, attention modules, and MLP is proposed, which achieves more accurate destination prediction while overcoming the structural deficiencies of MLP and convolution. The reliability of the CCMLP model is verified through different data and experiments. The experimental results show that selecting the trajectory features of the first m points and the last n points, as well as their corresponding direction angles, speeds, license plate numbers, and time features as the input features of the model can effectively improve the accuracy of destination prediction. The proposed CCMLP method has good feature learning ability. Compared with the MLP prediction model, the distance error is reduced by 10%, and compared with the ensemble learning algorithm, the distance error is reduced by 19.6%. The CCMLP model’s generalization ability for different data distributions is verified by dividing the data set based on weekdays and weekends, with distance losses of 2.25 km and 2.23 km respectively. The effect of different trajectory completeness on destination prediction is verified based on the first ten points of the trajectory, with distance losses of 2.23 km, 1.80 km, 0.97 km, and 0.68 km for completeness rates of 40%, 60%, 80%, and complete trajectories respectively.
    Reference | Related Articles | Metrics
    Improved A* Multi-Robot Bilevel Path Planning Algorithm
    CHEN Guangyou, YU Su
    Computer Engineering and Applications    2023, 59 (11): 312-319.   DOI: 10.3778/j.issn.1002-8331.2203-0197
    Abstract15)      PDF(pc) (601KB)(2)       Save
    In order to obtain the high-quality collision free path of multi-robots, a multi-robot bilevel programming algorithm based on improved A* algorithm and conflict coordination strategy is proposed. Firstly, in the first layer of the algorithm, the heuristic function of the traditional A* algorithm is improved by introducing the dynamic weight factor combined with planned path information to avoid its blind search, so as to speed up the search speed. Through the screening mechanism and Bezier curve, the actual turning times of the robot are reduced and planned path is smoothed, and then the initial path of a single robot is obtained. Secondly, in the second layer of the algorithm, the time dimension is introduced based on the two-dimensional path to establish the robot path time map, so as to predict the conflict between robots. Finally, the conflict coordination strategy is used to coordinate the conflicts between robots and between robots and dynamic obstacles. The experimental results show that the bilevel programming algorithm can effectively reduce planned path search time and the number of turns, and obtain the smooth and collision free navigation path of each robot.
    Reference | Related Articles | Metrics
    Multi-Objective Pick-Up Point Recommendation with Dynamic Constraint and Relaxed Segmentation
    GUO Yuhan, LIU Yuxi, LIU Qiuyue
    Computer Engineering and Applications    2023, 59 (11): 320-328.   DOI: 10.3778/j.issn.1002-8331.2203-0260
    Abstract14)      PDF(pc) (626KB)(4)       Save
    Optimization of online ride-sharing pick-up points effectively reduces driver pick-up and passenger walking distances, enhances pick-up efficiency, alleviates passenger aggregation and node congestion, and improves passenger travel experience. However, the results of the pick-up point planning are influenced by various factors, and it is challenging to address these factors comprehensively to ensure the feasibility of the planning scheme. In response to these challenges, a multi-objective integer linear planning model is proposed, covering passenger and driver benefit as well as road condition. Moreover, a relaxed segmentation iterative algorithm and a Pareto front generation method based on dynamic constraints are developed. Firstly, the spatial and temporal trajectory data is analyzed and filtered to obtain potential pick-up points that are accessible on the road network. Secondly, the relaxation problem is partitioned into subproblems by the branch-and-bound method, which is solved by the simplex method to convert the objective function to its standard form. Thirdly, the single-objective models are solved separately to obtain the boundary values, which are dynamically adjusted as constraints to generate Pareto surfaces. Finally, the correlation among the objectives is analyzed through a large number of experiments to derive the equilibrium point, so as to effectively recommend pick-up points from different perspectives.
    Reference | Related Articles | Metrics
    Multi-Task Banded Regression Model for Individual Cancer Survival Analysis
    WANG Huiheng, CAI Nian, CHEN Rui, LIU Xuan, LI Jian
    Computer Engineering and Applications    2023, 59 (10): 299-305.   DOI: 10.3778/j.issn.1002-8331.2201-0431
    Abstract28)      PDF(pc) (638KB)(17)       Save
    Individual cancer survival analysis is important to explore the survival status of a cancer patient. Most of previous multi-task linear regression methods based on linear transformation cannot reveal the cumulative risk of individual cancer course. A multi-task banded regression model(MTBR) is proposed for the survival analysis of the chronic and long-duration diseases such as cancer. Firstly, banded verification is introduced to design a multi-task regression model combining the regression and the banded transform. The transformation matrix for banded verification is mathematically deduced and verified in terms of validity and well-condition. Then, two specific matrices are presented as the examples of the multi-task banded matrix. Finally, the proposed model is validated by predicting the risk of individual survival analysis on the real-world cancer emergency and survival analysis datasets. The proposed model performs better survival analysis on the METABRIC dataset than the mainstream survival analysis models based on neural network, by the improvement of 0.05 C-index in 95% confidence intervals.
    Reference | Related Articles | Metrics
    Data Driven Optimal Combination Evaluation Model of Enterprise Credit Risk
    LUO Min, ZHOU Ligang, LIU Xinyue, ZHU Jiaming, CHEN Huayou
    Computer Engineering and Applications    2023, 59 (10): 306-313.   DOI: 10.3778/j.issn.1002-8331.2202-0261
    Abstract30)      PDF(pc) (673KB)(21)       Save
    Various types of information provided by corporate financial data can effectively explain the credit level of the company. However, too many indicators often have multicollinearity problems, resulting in the model overfitting and reducing the evaluation accuracy. In order to reduce the indicator data, the correlation coefficient between corporate default and financial indicators is calculated to eliminate indicators with weak correlation. The Lasso regression method is used to reduce the index data with high correlation. Then, three classification models of Logistic regression model, Bayesian model and support vector machine are used to classify and evaluate the credit risk of enterprises. Considering the different classification accuracy of different methods for different enterprises, in order to comprehensively utilize the advantages of each method, an optimal combination evaluation model of enterprise credit risk based on integer programming is constructed. Finally, the simulation analysis is carried out on the data of 300 companies listed on the Growth Enterprise Market. In order to verify the validity of the model, this paper randomly selects 3 groups of samples from the 300 companies(270 companies are training samples and 30 companies are test samples). The data of one year before the special treatment(ST) is performed are tested, and the experimental results show that the combined model has higher stability and classification accuracy.
    Reference | Related Articles | Metrics
    Research on Prediction Technology of Safety Index of Power Supply System Based on Deep Learning
    PENG Nan, GUO Jianfeng, ZHANG Wenxuan, WANG Jing, TAO Kai, HOU Senquan
    Computer Engineering and Applications    2023, 59 (10): 314-320.   DOI: 10.3778/j.issn.1002-8331.2202-0132
    Abstract27)      PDF(pc) (584KB)(21)       Save
    The safety index of high-speed railway power supply system reflects the occurrence trend of failures and accidents in the system, and the regular verification and prediction of the index are of very important practical significance for the comprehensive evaluation and pre-warning of the power supply system. Based on the data of the safety index over ten years, a time series predictive model is proposed, which combines time series statistics and deep learning methods. Firstly, ARIMA(autoregressive integrated moving average model) is used to model the time series of the safety index, and the adjust-R2 of the ARIMA model is improved by introducing the seasonal characteristic variable, verifying the seasonal law of the time series. Secondly, the GRU(gated recurrent unit) neural network is used to predict the time series. Finally, the Pearson coefficient is used to evaluate the effectiveness of the predictive model. The results show that the Pearson coefficients on the training set and test set are respectively 0.71 and 0.74, which proves the effectiveness of the predictive model.
    Reference | Related Articles | Metrics
    Research on GMV Prediction of E-commerce Based on LSTM-RELM Combination Model
    WANG Yiwen, WANG Weili
    Computer Engineering and Applications    2023, 59 (10): 321-327.   DOI: 10.3778/j.issn.1002-8331.2202-0067
    Abstract27)      PDF(pc) (733KB)(20)       Save
    With the development of the Internet, content marketing has gradually become the mainstream of e-commerce marketing, and the daily gross merchandise volume(GMV) is directly related to the inventory optimization control and advertising strategy of enterprises. In order to improve the prediction accuracy, based on the real e-commerce order data set, according to the content marketing index, the influence of user behavior on GMV is analyzed, and a combination model of long short-term memory network(LSTM) and regularized extreme learning machine(RELM) is proposed. The experimental results show more accurate prediction and faster running speed of the LSTM-RELM model proposed in this paper, compared with the traditional machine learning models and combination models, such as LSTM-LSTM, LSTM-SVR, GM(1, 1)-BPNN. The model can provide reference for advertising strategy and inventory optimization suggestions for relevant sales enterprises.
    Reference | Related Articles | Metrics
    Lightweight Semantic Segmentation Neural Network for Autonomous Driving
    XU Guobao, MAI Ruitao, YE Changxin, YAO Xu, LIU Mingxin
    Computer Engineering and Applications    2023, 59 (10): 328-334.   DOI: 10.3778/j.issn.1002-8331.2202-0092
    Abstract31)      PDF(pc) (686KB)(25)       Save
    Image semantic segmentation has very important applications in autonomous driving, allowing robots to segment semantic information in the environment to make decisions about downstream control actions. However, most of the deep learning models for this task are relatively large, require huge computing resources, and are difficult to use in mobile devices. In order to solve this problem, a lightweight neural network model for semantic segmentation is proposed, which uses a network architecture combining encoding-decoding and two-branch type. Grouping convolution, deep separable convolution, multi-scale feature fusion module and channel shuffling technology are used to reduce the number of network parameters and improve the prediction accuracy of the model. The model training in this paper combines Adam training method and stochastic gradient descent method. The Cityscapes data set is used, and 1000 training cycles are set. After testing, the number of model parameters is 3.5×106, and the calculation speed on a single graphics card GTX 1070Ti is 103 frames per second, which meets the real-time calculation standard. In the model evaluation indicators, the average intersection ratio is 61.3%, and the pixel accuracy rate is 93.4%, both of which are better than SegNet and ENet models.
    Reference | Related Articles | Metrics
    Application Method of Knowledge Graph Construction for UAV Fault Diagnosis
    QIU Ling, ZHANG Ansi, ZHANG Yu, LI Shaobo, LI Chuanjiang, YANG Lei
    Computer Engineering and Applications    2023, 59 (9): 280-288.   DOI: 10.3778/j.issn.1002-8331.2210-0415
    Abstract46)      PDF(pc) (713KB)(31)       Save
    In recent years, the safety and security of UAV operations have faced serious challenges, and it is crucial to ensure safe UAV operations. Based on the emerging research hotspot of knowledge mapping, this paper can make full use of UAV a priori knowledge for fault diagnosis, which can realize component association diagnosis and achieve interpretability of diagnosis results by relying on expert knowledge. At present, there are few studies on knowledge graphs for fault diagnosis, and usually“pre-training”models are used to solve the problem of insufficient data for deep learning model training. Still, the application scenarios of this method are more restricted and cannot provide valuable reference training samples for subsequent researchers. Based on the UAV fault repair manual as the primary data, a remotely supervised data annotation method based on the human school of the aircraft standard is proposed to obtain a substantial and accurately annotated UAV fault corpus, and the rule-based and BiLSTM-CRF network knowledge extraction method is combined according to the data structure characteristics, and the experiments prove that the entity extraction effect is good. Based on the UAV fault diagnosis ontology, this paper completes the construction of the UAV fault diagnosis knowledge graph, stores and visualizes it through Neo4j, and builds an intelligent question and answer system for UAS faults to provide a reasoned and accurate diagnosis for UAV faults, which proves the effectiveness of the knowledge graph in the fault diagnosis field and provides a scientific basis for the construction of the fault diagnosis system based on the knowledge graph.
    Reference | Related Articles | Metrics
    UAV Cluster Forest Fire Detection Method Based on PSO-GA Algorithm
    HUANG Zhibin, CHEN Xun
    Computer Engineering and Applications    2023, 59 (9): 289-294.   DOI: 10.3778/j.issn.1002-8331.2201-0092
    Abstract50)      PDF(pc) (603KB)(39)       Save
    In the field of aerial forest protection, the application of UAVs for forest fire detection suffers from the dilemma of low detection efficiency and low intelligence. By taking advantage of the smoke spreading characteristics of forest fires and combining the advantages of UAV clusters, the UAV clusters can conduct intelligent fire detection according to the smoke situation at the fire site, and at the same time, the UAV cluster detection algorithm is optimized in three aspects:linear reduction of inertia weights, fusion algorithm and forest wind characteristics. This paper proposes PSO-GA algorithm that combines the features of particle swarm algorithm and genetic algorithm. The proposed particle swarm genetic algorithm avoids the algorithm to fall into local optimum, and improves the efficiency and stability of UAV cluster for forest fire detection. The simulation results prove the effectiveness of forest fire detection by UAV cluster based on smoke concentration, and the experimental results show that PSO-GA algorithm has better optimality and convergence than traditional particle swarm algorithm and fish swarm algorithm, and shortens the detection time of forest fire. The above research can provide effective support for forest fire detection and prevent the spread of forest fires effectively.
    Reference | Related Articles | Metrics
    Joint Optimization Method for Order Split and Delivery Based on Multi-Store Collaboration
    ZHANG Yanju, OU Liping
    Computer Engineering and Applications    2023, 59 (9): 295-303.   DOI: 10.3778/j.issn.1002-8331.2201-0201
    Abstract39)      PDF(pc) (633KB)(24)       Save
    New retail has led to the transformation of traditional enterprises, resulting in the emergence and continuous development of order fulfilment models in which physical stores act as front warehouses. In response to order demand uncertainty and store inventory changes, the problem of joint optimization of order split and delivery under multi-store collaboration is proposed for situations arising in order fulfilment at the nearest store. By introducing a limit on the number of split orders, the problem-solving space is reduced. To reduce the path overlap caused by separate deliveries, a collaborative delivery model is used to integrate the paths. The order fulfilment cost is reduced by optimizing the adjustment between order split and delivery. Integrating breadth-first search and local search algorithms, the TNILS hybrid heuristic algorithm is constructed to solve the problem. Based on the synthetic dataset, the effectiveness of collaborative delivery and the feasibility of the proposed algorithm are demonstrated by comparing the results of collaborative delivery with those of separate delivery. Finally, the effectiveness and stability of the TNILS(top-N & improved local search)algorithm are verified by comparing the experimental results with other algorithms.
    Reference | Related Articles | Metrics
    Research on Knitted Production Line Scheduling Based on Improved Simulated Annealing Algorithm
    DU Lizhen, WANG Yuhao, XUAN Zifeng, YE Tao, ZHANG Yajun
    Computer Engineering and Applications    2023, 59 (9): 304-312.   DOI: 10.3778/j.issn.1002-8331.2201-0315
    Abstract61)      PDF(pc) (778KB)(25)       Save
    In order to solve the scheduling problem of knitting garment production line with multi-stage production flow, a four-stage heterogeneous shop scheduling model with machine resource constraints and batch constraints is firstly constructed, and multiple scheduling rules are used to connect the scheduling flow of adjacent stages. Then, an improved simulated annealing algorithm based on slice sorting with multi-neighborhood search and reheating operation is proposed to solve the above model with the goal of the minimum makespan. Finally, combined with the actual background of the enterprise, 10 kinds of hybrid orders are constructed as examples for simulation experiments, and compared with the existing optimization algorithms, the effectiveness of the algorithm in solving the scheduling problem of knitted garment production line is verified.
    Reference | Related Articles | Metrics
    Optimal Strategy of Differential Game Pursuit Problem in Graph Attention Network
    LIU Zhaolong, SONG Yao, XU Yiming, FAN Xinyue
    Computer Engineering and Applications    2023, 59 (9): 313-318.   DOI: 10.3778/j.issn.1002-8331.2201-0148
    Abstract39)      PDF(pc) (619KB)(11)       Save
    The optimal strategy for the pursuit of fugitives in differential game is based on the trajectory prediction model of the pursuit of fugitives, and the prediction is made through the trajectory of both parties, so as to make a more predictable dynamic strategy.?Therefore, in order to obtain the optimal strategy of both sides in the game, the algorithm of random motion of both sides is proposed and designed, and the state equation of chasing both sides is established. On this basis, the adjacency matrix and the connection mode of feature data in the network are redesigned by improving the graph attention network(GAT).?The trajectory prediction model of attacker and target is constructed and verified numerically.?In addition, the method of covering the trajectories of random motion by ring is used to establish the trajectory connection graph.??The results show that GAT network is superior to graph convolution network and Chebshev spectrum convolution network in MAE, MAPE and RMSE, and can be used to study the optimal strategy of differential game pursuit problem.
    Reference | Related Articles | Metrics
    Graph Convolutional Index Trend Prediction Based on Correlation of Index Constituent Stocks
    WANG Changhai, LIANG Hui, WANG Bo, CUI Xiaoxu
    Computer Engineering and Applications    2023, 59 (9): 319-328.   DOI: 10.3778/j.issn.1002-8331.2205-0270
    Abstract46)      PDF(pc) (906KB)(16)       Save
    Predicting future trends of stock market indexes with historical transaction data is an important issue in the field of financial investment. Fusing the trends correlations between indices with the graph convolutional network is a research hotspot. Aiming at the inconsistency of historical and future dynamic graph structures in the current graph convolutional index prediction, a graph convolutional index trend prediction method termed G-Conv is proposed, in which the constituent stocks are applied to construct the index graph. The traditional quantitative features and deep features of one-dimensional convolutional networks are extracted as the features of predict samples. Then, the graph between indices is constructed using the constituent data of the index, and graph convolution is performed on features of different index samples to obtain the index prediction result. Finally, 42 commonly used indices in the A-shares market are used to evaluate the performance of this method. In the experiment, MAE and MSE are used as loss functions of model training, and classical methods such as GC-CNN, AD-GAT are selected as comparison benchmarks. The results show that G-Conv reduces the average prediction error by 5.10% and 4.20% under these two error evaluation criteria respectively, and shows better generalization performance.
    Reference | Related Articles | Metrics
    3D Object Detection in Substation Scene Based on Graph Neural Network
    ZHANG Ting, ZHANG Xingzhong, WANG Huimin, YANG Gang, WANG Dawei
    Computer Engineering and Applications    2023, 59 (9): 329-336.   DOI: 10.3778/j.issn.1002-8331.2111-0264
    Abstract63)      PDF(pc) (728KB)(31)       Save
    In the three-dimensional scene of a substation,the precise positioning and identification of inspectors and live equipment is a prerequisite for improving the level of personnel safety management and control. Aiming at the problem of inaccurate target positioning and recognition in complex scenes of substations, a method of 3D object detection in substation scene based on graph neural network is proposed. The method is designed based on the point-GNN structure. In the vertex feature extraction stage, the PCS(point-channel-sphere) attention structure is proposed to extract more abundant key point feature information. In the GNN edge feature aggregation stage, an overall pooling mechanism is adopted to take into account the maximum pooling and the mean pooling to obtain richer global features, improving the model loss function, using Focal Loss as the classification loss to make the training pay more attention to the previous scenic spots, and using DIoU Loss as the regression loss to make the regression task more efficient. Training and testing on the self-built substation scene dataset, experiments show that the mAP value of this method reaches 73.81%, which is better than the benchmark model. It can improve the detection effect of objects in the substation scene and has certain practical value for improving the level of personnel safety management and control.
    Reference | Related Articles | Metrics
    Mars Unmanned Aerial Vehicles Control with Deep Deterministic Policy Gradient
    SUN Dan, ZHENG Jianhua, GAO Dong, HAN Peng
    Computer Engineering and Applications    2023, 59 (8): 288-296.   DOI: 10.3778/j.issn.1002-8331.2112-0528
    Abstract40)      PDF(pc) (814KB)(16)       Save
    In order to reduce the dependence of controller design on Mars unmanned aerial vehicle(UAV) dynamic models and improve the intelligence level of Mars UAV control system, a reinforcement learning-based controller for Mars UAV is proposed. The controller consists of neural networks and is trained by deep deterministic policy gradient(DDPG) algorithm. Finally, it obtains a control strategy to meet the control requirements according to current states and targets. The simulation results demonstrate that the controller based on DDPG is able to control the Mars UAV to a specified position autonomously without the derivation of UAV dynamic model. Mean-while, the performance such as control precision and adjustment time reaches the effect of proportion integration differentiation (PID) controller, which verifies the effectiveness of DDPG-based controller. In addition, when the controlled object model changes or there is external disturbance, the controller based on DDPG still completes the task stably, and the control effect is better than PID controller, indicating that the controller based on DDPG has good robustness.
    Reference | Related Articles | Metrics
    Feature Mining for Financial Texts and Research on Dynamic Factor Fusion Strategy
    ZHANG Wei, ZHU Hanqing, GAO Zhigang
    Computer Engineering and Applications    2023, 59 (8): 297-305.   DOI: 10.3778/j.issn.1002-8331.2112-0142
    Abstract59)      PDF(pc) (620KB)(18)       Save
    Current analysis of financial texts is limited by non-normative financial texts, and the extracted financial features are not effective enough. To solve this problem, an normative finanical text feature mining(NFTFM) model that studies brokerage research reports is proposed to extract valid financial feature factors from normative financial texts. Firstly, the normative finanical text sentiment dictionary(NFTSD) is proposed to fully mine the semantics of brokerage reporting. Then, it uses the [K]-nearest neighbor algorithm(KNN) to classify the attitude tendency of the report authors. Finally, it integrates the attitude classification results into two financial characteristic factors which include rate consistency(RV) factor and rate consistency(RC) factor according to the temporal dimension. Aiming at the problem that the factor weight of traditional quantitative multi-factor model can not adapt to the market change, the fusion factor strategy of dynamic optimization is proposed, and the weight of the factors are dynamically optimized by genetic algorithm. In order to verify the effectiveness of normative financial characteristic factors and the effect of dynamic optimization fusion factor strategy, the RC and RV factors are taken as the basic factor set to construct a multi-factor strategy instance for China Securities 500 stocks and carry out historical cycle backtest. The results show that the strategy returns have significantly improved compared with the benchmark returns, and have good adaptability to different market environments, indicating that NFTFM model effectively extracts standardized financial feature factors, and the factors under the dynamic optimization of the fusion factor strategy have the ability to adapt to market changes.
    Reference | Related Articles | Metrics
    Application of SAC-Based Autonomous Vehicle Control Method
    NING Qiang, LIU Yuansheng, XIE Longyang
    Computer Engineering and Applications    2023, 59 (8): 306-314.   DOI: 10.3778/j.issn.1002-8331.2112-0084
    Abstract53)      PDF(pc) (939KB)(39)       Save
    In order to improve the problem of slow network convergence and unstable training process caused by equal probability sampling of SAC(soft actor critic) algorithm samples and random initialization of the network, an improved algorithm PE-SAC(priority playback soft actor) is proposed that combines priority playback and expert data. The algorithm classifies the sample pool according to the sample value, uses expert data to pre-train the network, reduces the invalid exploration space of unmanned vehicles, reduces the number of trials and errors, and effectively improves the learning efficiency of the algorithm. At the same time, a reward function for multiple obstacles is designed to enhance the applicability of the algorithm. Simulation experiments are carried out on the CARLA platform, and the results show that the proposed method can better control the safe driving of unmanned vehicles in the environment, and the reward value and convergence speed obtained under the same training times are better than TD3(twin delayed deep deterministic policy gradient algorithm) and SAC algorithm. Finally, combined with the radar point cloud map and the PID(proportional integral derivative) control method, the difference between the simulation environment and the real scene is reduced, and the training model is transplanted to the low-speed unmanned vehicle in the park to verify the generality of the algorithm.
    Reference | Related Articles | Metrics
    Optimal Maintenance Decision Model for Sensor Network
    LI Shaohua, GUO He, FENG Jingying, HE Wei
    Computer Engineering and Applications    2023, 59 (8): 315-321.   DOI: 10.3778/j.issn.1002-8331.2112-0362
    Abstract33)      PDF(pc) (694KB)(15)       Save
    Optimal maintenance decision for a sensor network aims to provide the optimal repair time for users. The accuracy of the optimal maintenance decision method directly determines the reliability of the sensor network. In this paper, a new optimal maintenance decision model based on belief rule base(BRB) expert system is developed to solve two problems:the lack of observation data and complex system mechanism. In the new model, two parts exist:the health state assessment and the health state prediction. First, the health state of the sensor network is estimated by a BRB based health assessment model. Then, based on the current health state, a Wiener process is used to predict the health state of the sensor network, and the optimal maintenance time can be obtained. In the Wiener process-based health state prediction model, the minimum threshold of the health state of the sensor network should be provided by experts to determine the optimal maintenance time. To illustrate the effectiveness of the proposed model, a case study for the wireless sensor network of oil storage tank is conducted.
    Reference | Related Articles | Metrics
    Short-Term Passenger Flow Prediction of Airport Subway Based on Spatio-Temporal Graph Convolutional Network
    ZHANG Xingrui, LIU Chang, CHEN Zhe, DENG Qiangqiang, LYU Ming, LUO Qian
    Computer Engineering and Applications    2023, 59 (8): 322-330.   DOI: 10.3778/j.issn.1002-8331.2112-0383
    Abstract57)      PDF(pc) (739KB)(40)       Save
    The short-term passenger flow prediction of the airport subway is the key to realize the rapid evacuation of airport passengers and the command and dispatch of on-site capacity resources of the terminal. Taking into account the complex spatial structure of the airport and the impact of flight fluctuations, a short-term passenger flow prediction model for airport subways which combined graph convolutional neural network(GCN) and gated convolution(GLU) is established. Based on the graph convolutional neural network(GCN), the spatial structure relationship between airport path points and subway gate is obtained. At the same time, a combined gated convolution method is proposed to mine the time-varying characteristics of subway passenger flow under flight fluctuations, and effectively capture the volatility of subway passenger flow. Taking the Capital Airport T3 terminal as the research object, after many experiments, the root mean square error, average absolute error and average absolute percentage error of the prediction performance of the graph convolution spatio-temporal prediction model are all smaller than the traditional ARIMA prediction model and the LSTM, STGCN model. The results show that the model can capture the fluctuation relationship between subway passenger flow and flight passenger flow, has high prediction accuracy, and improves the robustness of model prediction.
    Reference | Related Articles | Metrics
    Non-Contact Atrial Fibrillation Detection Based on Video Pulse Features
    ZHANG Xu, YANG Xuezhi, LIU Xuenan, FANG Shuai
    Computer Engineering and Applications    2023, 59 (8): 331-340.   DOI: 10.3778/j.issn.1002-8331.2201-0385
    Abstract37)      PDF(pc) (713KB)(8)       Save
    Early detection of atrial fibrillation is very important for the prevention of cardiovascular and cerebrovascular diseases. This paper proposes a facial video atrial fibrillation detection method. The method extracts the pulse signals from the facial video by face tracking and improved complete ensemble empirical modes decomposition(ICEEMD), and extracts atrial fibrillation discriminative features from video pulse signals according to the pulse characteristics during atrial fibrillation episodes. An improved recursive feature elimination feature selection method is designed to screen out the more important features for atrial fibrillation detection. Based on the above features, machine learning methods are used to achieve atrial fibrillation detection. Experiments are conducted on 122 cases of atrial fibrillation patients and 139 cases of normal sinus rhythm facial videos. The optimal features are the percentage of heartbeats with adjacent RR interval greater than 50 ms (PNN50), the maximum value of RR intervals (maxRR), and the horizontal radius of poincare diagram (SD2) etc. Based on the above optimal feature set, the accuracy of atrial fibrillation detection is 92.31%, the specificity is 90.24%, the sensitivity is 94.59%, and the AUC is 0.920 5.
    Reference | Related Articles | Metrics
    Research on Online Selling Format Choice and Cooperative Advertising of Manufacturer
    CHEN Weiyu, WEI Jie, ZHAO Shuo
    Computer Engineering and Applications    2023, 59 (7): 269-277.   DOI: 10.3778/j.issn.1002-8331.2207-0079
    Abstract51)      PDF(pc) (564KB)(24)       Save
    Based on a supply chain composed of a manufacturer and an online retailer, this paper studies the interaction between selling format and cooperative advertising. This paper firstly develops four Stackelberg game models under different selling formats and different cooperative advertising strategies by considering the common impact of retailing price and advertising level on demand, and then compares channel members’ profits under different models. The results show that under agency format, cooperative advertising can boost both the manufacturer’s and the online retailer’s profit. However, under reselling format, the manufacturer tends to bear less than half of the advertising expenses, and this behavior will hurt the online retailer’s profit. Besides, the manufacturer should choose agency format regardless of the cooperative advertising strategy. The online retailer should choose reselling format under non-cooperative advertising strategy, while its choice for selling format depends on the advertising cost share rate under cooperative advertising strategy.
    Reference | Related Articles | Metrics
    Research on End-To-End Nested Named Entity Recognition Method
    DENG Liyuan, CHEN Yanpin, WU Yuefei, QIN Yongbin, HUANG Ruizhang, ZHENG Qinghua, TAN Xi
    Computer Engineering and Applications    2023, 59 (7): 278-284.   DOI: 10.3778/j.issn.1002-8331.2109-0498
    Abstract58)      PDF(pc) (621KB)(31)       Save
    Named entity recognition(NER) is regarded as a basic research in natural language processing. Inspired by the one-stage object detection algorithm in computer vision, this paper proposes an effective end-to-end method for identifying nested named entities by using its algorithm idea and introducing regression operation. Based on multi-task learning framework, the paper uses deep neural network to transform sentences into text feature graphs to regress nested entity boundaries, and designs centrality method to suppress low-quality boundaries. A comparative experiment is carried out with several methods on ACE2005 Chinese dataset. Experiments show that this method is effective in identifying nested named entities in text, and the computer vision algorithm idea and boundary regression mechanism achieve ideal results in natural language processing tasks.
    Reference | Related Articles | Metrics
    Rail Surface Defect Method Based on Bimodal-Modal Deep Learning
    ZHAO Hongwei, ZHENG Jiajun, ZHAO Xinxin, WANG Shengchun, LI Yidong
    Computer Engineering and Applications    2023, 59 (7): 285-293.   DOI: 10.3778/j.issn.1002-8331.2209-0364
    Abstract83)      PDF(pc) (781KB)(21)       Save
    Aiming at the problem that the current rail surface detection system lacks the third dimension key depth information and the detection results are vulnerable to interference with high false alarm rate, the paper designs a rail surface defect detection system and method based on dual mode structured light sensor. By constructing a multi-mode depth learning detection network for rail surface defects, the defects in the bimodal rail images can be detected. The depth network proposed fuses multi-scale features of bimodal images respectively, and conducts multi-scale rail head surface defect detection. Experimental results show that this method can significantly reduce false positives while maintaining a high detection rate. Compared with the common depth learning detection model in the current defect detection, the mean average accuracy(mAP) has been greatly improved, and its performance is better than the previous detection algorithm, which has a good application prospect in the rail head surface scratch detection task.
    Reference | Related Articles | Metrics
    Online Car-Hailing Dispatch Method Based on Local Position Perception Multi-Agent
    HUANG Xiaohui, LING Jiahao, ZHANG Xiong, XIONG Liyan, ZENG Hui
    Computer Engineering and Applications    2023, 59 (7): 294-301.   DOI: 10.3778/j.issn.1002-8331.2111-0490
    Abstract41)      PDF(pc) (666KB)(16)       Save
    In recent years, online car-hailing has become an indispensable part of people’s daily travel. The core task of the online car-hailing platform is how to effectively dispatch the order to the appropriate driver, so that the overall waiting time of users is as short as possible, and the driver’s revenue is as high as possible. In the current research, greedy algorithms and reinforcement learning are mainly used to build models. However, current methods mostly only consider the immediate satisfaction of passengers, and fail to effectively consider the relative position relationship between vehicles and orders, and reduce the waiting time of all passengers from a long-term perspective. For this reason, this paper constructs order dispatch as a Markov process, and proposes a multi-agent vehicle dispatch method based on local position perception. This method captures the space-time relationship between people and vehicles by designing appropriate input states and convolutional neural networks, and reduces the overall waiting time of passengers from a long-term perspective. Experimental results show that in scenarios with different specifications of maps, different numbers of vehicles and orders, the method proposed is superior to existing methods and has better generalization capabilities. Especially in large-scale human-vehicle environments. the results obtained by the method are significantly better than the existing methods.
    Reference | Related Articles | Metrics
    Improved Constrained Artificial Bee Colony Algorithm and Its Financial Application
    ZHI Junyang, WANG Zhen, CUI Keke
    Computer Engineering and Applications    2023, 59 (7): 302-310.   DOI: 10.3778/j.issn.1002-8331.2111-0031
    Abstract43)      PDF(pc) (583KB)(24)       Save
    An improved artificial bee colony algorithm is proposed for constrained optimization problems. The Pareto dominate criterion is introduced to improve the exploration ability of the algorithm and avoid the premature convergence of the algorithm. In the employed bee stage, the searching equations and constraint handling strategies are selected adaptively by identifying the current state of the population, which can guide the population into the feasible region quickly. In the onlooker bee stage, the global optimal solution is used for guiding the population searching, which can improve the exploitation ability of the algorithm. Experimental results on 20 benchmark test functions in CEC 2006 show that this algorithm can effectively solve constrained optimization problems. Furthermore, the portfolio optimization problems are solved by this algorithm. The numerical experiments show the effectiveness of this algorithm for portfolio optimization problems, which can solve this kind of financial problem effectively.
    Reference | Related Articles | Metrics
    Prediction of Fault Symptoms in Elevator Knowledge Base Based on Improved LSTM-AE Algorithm
    SUN Qinggang, WANG Cheng
    Computer Engineering and Applications    2023, 59 (7): 311-318.   DOI: 10.3778/j.issn.1002-8331.2111-0134
    Abstract50)      PDF(pc) (649KB)(19)       Save
    To solve the problem of fault symptom prediction in the operation and maintenance knowledge base system, an improved LSTM-AE algorithm for elevator equipment is proposed. Firstly, in view of redundancy problem in the elevator operating parameter sequence, the attribute correlation density ranking(ACDR) method is used to filter the feature vector group. Moreover, aiming at the non-stationary problem of the running speed characteristic sequence, variational mode decomposition(VMD) algorithm is used for noise reduction and smoothing. Finally, a sliding window attention mechanism fused with BILSTM is introduced into the LSTM-AE model to improve the ability of time-series feature extraction, and softmax classifier is used to fuse the reconstruction error of each feature sequence to realize prediction of the elevator fault symptom. The experimental results show that compared with traditional LSTM-AE algorithm, the proposed improved LSTM-AE algorithm has a 13% increase in the normal sample accuracy rate and a 11% reduction in the abnormal sample error rate. It can predict common elevator faults more precisely, and be suitable for constructing a reliable fault symptom prediction model of elevator operation and maintenance knowledge base.
    Reference | Related Articles | Metrics
    Research on Two-Stage Pricing of Shipping Supply Chain Under Blockchain Platform
    WANG Xiaoguang, YIN Meng
    Computer Engineering and Applications    2023, 59 (7): 319-327.   DOI: 10.3778/j.issn.1002-8331.2208-0044
    Abstract39)      PDF(pc) (629KB)(19)       Save
    Considering the decentralized technology of blockchain and the pricing decisions of members in the shipping supply chain, a port-carrier secondary shipping supply chain is constructed. The port establishes a private blockchain platform for information sharing, combined with the information sharing degree coefficient and the cost sharing ratio coefficient, constructs the pricing decision-making model in the traditional model and the blockchain technology-based model, and compares the optimal pricing strategies of the two models using Stackelberg game. The impact of different degrees of information sharing on private blockchain platforms on pricing and revenue is discussed. The results of the study found that after the port establishes a private blockchain platform, regardless of the degree of information sharing, it can always increase the carrier’s freight wholesale price and the port wholesale price. When the information is fully shared, the port has the greatest benefits, but at this time the carrier needs to bear all the platform construction costs, but its own revenue will be reduced and give up joining the private blockchain platform, in order to develop the carrier that can appropriately bear a certain proportion of the platform cost, under certain conditions can reach the optimal strategy.
    Reference | Related Articles | Metrics
    Research on Optimization of Production Sequence of Multi-District Parallel Picking System in Distribution Center
    AN Yuxin, WANG Zhuan
    Computer Engineering and Applications    2023, 59 (7): 328-336.   DOI: 10.3778/j.issn.1002-8331.2111-0042
    Abstract51)      PDF(pc) (611KB)(19)       Save
    With the rapid development of e-commerce industry, picking operation system and process of distribution center become more and more complex. When batch orders are issued, how to shorten the completion time as much as possible under the premise of reducing queuing has become a key problem for enterprises to improve picking efficiency and reduce logistics costs. Considering that most of the current picking systems in large distribution centers adopt multi-zone parallel picking strategy, and the order process of different structures is different, the two-layer optimization objective is to minimize the completion time and the waiting time of collection single queue, aiming at the whole process of picking system from picking to packing out of the warehouse. An optimization model of collection single production sequence of picking system based on multi-zone parallel picking is established. To solve the multi-objective problem, a multi-objective solution method based on fast non-dominated sequencing genetic algorithm is designed, and digital simulation method is introduced to calculate the fitness value. Finally, the effectiveness of the algorithm is proved by empirical analysis. The results show that the optimization method has good practical value for improving the operation efficiency of picking system in distribution center.
    Reference | Related Articles | Metrics
    Research on Event Extraction Method for Judicial Data
    JIA Zhen, DING Zehua, CHEN Yanping, HUANG Ruizhang, QIN Yongbin
    Computer Engineering and Applications    2023, 59 (6): 277-282.   DOI: 10.3778/j.issn.1002-8331.2109-0497
    Abstract51)      PDF(pc) (909KB)(31)       Save
    The events in the judicial data are mainly used to describe the changes in the behavioral state between the criminal subject and the object in the case, and the identification of judicial events can effectively support the intelligent auxiliary case-handling research. At present, the existing event extraction technology mainly recognizes events through trigger words, and then extracts corresponding parameters according to a predefined template. The disadvantage of it is that only predefined event types can be extracted, and the extracted events are not necessarily the center of the sentence semantic expression. Aiming at the above problems, this paper proposes a method for defining judicial events based on the predicate head, and builds a neural network model that combines the meaning of words and words. The model uses embedding of words to obtain semantic information of words, and obtains word feature information through CNN. After combining the word feature information, Cross-BiLSTM is used to cross-learn the context-dependent representation of the word interaction information. Finally, the CRF calculates the optimal tag path for each word. Experiments show that the F1 value of the model on the judicial data set reaches 84.41%, which exceeds the comparison method by 4.8%.
    Reference | Related Articles | Metrics
    Constrained Multi-Start Path Planning Algorithm on Large-Scale Graphs
    PU Linfa, YANG Yajun, WANG Xin
    Computer Engineering and Applications    2023, 59 (6): 283-290.   DOI: 10.3778/j.issn.1002-8331.2109-0496
    Abstract52)      PDF(pc) (645KB)(30)       Save
    Path planning query is a basic problem on graph data, which has important application value in many fields. Usually, the path queried in actual problems is constrained. For example, in the problems of food delivery and shared travel, the path has node constraints, and the path needs to meet the constraints of the sequence relationship between nodes. At present, for the path query problem with node constraints, most of the work is on the single-start node-constrained path query, but it is difficult to extend to the multi-start node constraint problem. Because the multi-start path query problem with node constraints is NP-hard, most of the existing methods for this problem use greedy incremental processing, but it is insufficient for processing static rule sets. Therefore, a heuristic algorithm based on sub-paths and an accurate algorithm based on constraint set expansion are proposed, and the effectiveness of the algorithm is verified on the real data set. The experimental show that the heuristic algorithm can give an accurate solution to the problem, and the heuristic algorithm can quickly give a better approximate solution.
    Reference | Related Articles | Metrics
    Tracking and Registration Method Based on Point Cloud Matching for Augmented Reality Facing Work System
    JIA Xiaohui, FENG Chongyang, LIU Jinyue
    Computer Engineering and Applications    2023, 59 (6): 291-298.   DOI: 10.3778/j.issn.1002-8331.2110-0379
    Abstract41)      PDF(pc) (653KB)(15)       Save
    This paper proposes a tracking and registration method based on point cloud matching of augmented reality to solve the occlusion problem in facing work of construction robot. Firstly, the point cloud of the target model and the operating environment is matched and the location of target is initialized. Secondly, the improved correlation filter tracking algorithm is adopted to track the target and obtain the target position. Then, the ICP(iterative closest point) method is applied to estimate the pose of target. Finally, pose optimization is added in the tracking registration process. In order to track the target more accurately, the scale-adaptive correlation filter tracking based on multiple features is proposed. Through the panel installation experiments are carried out and the results show that the proposed method has good accuracy and real-time performance. The minimum positioning error is 2.88 mm, and the effect of virtual and reality fusion is perfect.
    Reference | Related Articles | Metrics
    Chimp Optimization Algorithm Improved by Chaos Elite Pool Collaborative Teaching-Learning and Its Mechanical Application
    LUO Shihang, HE Qing
    Computer Engineering and Applications    2023, 59 (6): 299-309.   DOI: 10.3778/j.issn.1002-8331.2110-0273
    Abstract49)      PDF(pc) (928KB)(32)       Save
    Chimp optimization algorithm improved by the elite chaos pool collaborative teaching-learning is proposed overcome the drawbacks of weak global search ability, low optimization accuracy, and slow convergence speed. This paper uses the chaotic elite pool strategy to generate the initial population, enhances the quality of the initial solution and the diversity of the population, and lays the foundation for the global optimization of the algorithm. Furthermore, adaptive oscillation factors are introduced to balance ChOA’s global exploration and local development capabilities. Finally, combining the teaching phase of the teaching and learning optimization algorithm and the individual memory idea of the particle swarm optimization algorithm to optimize the population position update process, and improve the algorithm’s optimization accuracy and convergence speed. The simulation experiment compares ECTChOA with standard ChOA, other meta-heuristic optimization algorithms, and the latest improved ChOA under 12 benchmark functions. The experimental results and the Wilcoxon rank sum test [p]-value results both show that the improved algorithm has higher search accuracy, faster convergence speed and better robustness. In addition, ECTChOA is applied to mechanical engineering design cases to further verify the feasibility and applicability of ECTChOA in actual engineering problems.
    Reference | Related Articles | Metrics
    Surface Crack Detection in Ballastless Slab Track of High-Speed Railway Based on Improved RetinaNet
    ZHANG Shihui, LUO Hui, PEI Yingling, YU Junying, XU Jie
    Computer Engineering and Applications    2023, 59 (6): 310-317.   DOI: 10.3778/j.issn.1002-8331.2110-0419
    Abstract63)      PDF(pc) (791KB)(39)       Save
    A crack detection method based on improved RetinaNet is proposed to address the problems of large difference between scales and unbalanced crack types in the surface of ballastless slab track of high-speed railway. To alleviate the problem of subtle information loss caused by downsampling and horizontal connection compression of feature pyramid, multi-level feature pyramid network is used to integrate different depth features extracted from ResNet-50 backbone network to achieve full expression of image feature information. To solve the problem of mismatching between the classification and positioning confidence of surface cracks in the detection process, adaptive anchor learning is proposed to optimize the anchor and the network model at the same time, which improves the detection accuracy of small-scale cracks. To alleviate the impact of crack category imbalance in detection performance, Focal Loss function is introduced as the classification loss function, and weight factor of class balance is added to improve the detection accuracy of small types of cracks. The experimental results show that the improved RetinaNet detection network achieves good results on different crack types in the ballastless slab track of high-speed railway, and the mean average precision(mAP) is 72.58%, which is 3.60 percentage points higher than that of the original RetinaNet detection network, effectively realizes the accurate detection of cracks of different scales.
    Reference | Related Articles | Metrics
    Research on Motion Control Method of Manipulator Based on Reinforcement Learning
    YANG Bo, WANG Kun, MA Xiangxiang, FAN Biao, XU Lei, YAN Hao
    Computer Engineering and Applications    2023, 59 (6): 318-325.   DOI: 10.3778/j.issn.1002-8331.2207-0159
    Abstract45)      PDF(pc) (690KB)(20)       Save
    The traditional motion control algorithm has the problems of poor environmental adaptability and low efficiency. Reinforcement learning can be used to constantly explore trial and error in the environment, and the motion of the manipulator can be controlled by adjusting the neural network parameters through the reward function. However, in reality, it is impossible to provide a trial and error environment for the manipulator. This paper uses the Unity engine platform to build a digital twin simulation environment for the manipulator, set the observation state variables and set the reward function mechanism, and proposes the M-PPO algorithm combining PPO(proximal policy optimization) and multi-agent(agents) in this model environment to speed up the training speed and realize intelligent motion control of the manipulator through reinforcement learning algorithms. This paper completes the effective obstacle avoidance at the end of the manipulator’s execution and reach the target object’s position quickly, and also analyzes the experimental results of the algorithm, M-SAC(multi-agent and soft actor critical) and PPO algorithm. The effectiveness and progressiveness of M-PPO algorithm is verified in the debugging of the manipulator’s motion control decision under different environments. It achieves the purpose of independent planning and decision-making of twins and reverse control of synchronous movement of physical bodies.
    Reference | Related Articles | Metrics