In response to the increasing complexity of Industrial Control System （ICS） structure， especially within the context of cloud-edge collaborative computing， which significantly raises cybersecurity risks， an evaluation framework specifically for assessing cross-domain attacks in cloud-edge collaborative scenarios was proposed to identify， evaluate， and defense against potential security threats systematically. Initially， this framework entailed a thorough collection and categorization of ICS assets， cross-domain attack entrances， methods， and impacts， establishing a foundational database and structure for assessment. Furthermore， based on the characteristics of ICS， a novel set of evaluation indicators for cross-domain attacks was developed， encompassing system modules， attack paths， attack methods， and potential impacts. Additionally， through simulation experiments conducted in a simulated ICS environment， the effectiveness of this evaluation framework was tested， verifying its capacity to effectively identify vulnerabilities within the system and enhance overall security. The results demonstrate that the assessment framework can provide both theoretical and practical guidance for the secure application of cloud-edge technologies in industrial settings， indicating promising applicability.

To address the issue of insufficient extraction of semantic feature information with different scales and the lack of focus on crucial information when obtaining sentence semantic information by Convolutional Neural Network （CNN）-based relation extraction， a model for relation extraction based on a multi-scale hybrid attention CNN was proposed. Firstly， relation extraction was modeled as label prediction with two-dimensional representation. Secondly， by extracting and fusing multi-scale feature information， finer-grained multi-scale spatial information was obtained. Thirdly， through the combination of attention and convolution， the feature maps were refined adaptively to make the model concentrate on important contextual information. Finally， two predictors were used jointly to predict the relation labels between entity pairs. Experimental results demonstrate that the multi-scale hybrid convolutional attention model can capture multi-scale semantic feature information，And the key information in channels and spatial locations was captured by the channel attention and spatial attention by assigning appropriate weights， thereby improving the performance of relation extraction. The proposed model achieves F1 scores of 90.32% on SemEval （SemEval-2010 task 8） dataset， 70.74% on TACRED （TAC Relation Extraction Dataset）， 85.71% on Re-TACRED （Revised-TACRED）， and 89.66% on SciERC （Entities， Relations， and Coreference for Scientific knowledge graph construction）.

Existing methods for answer acquisition based on pre-trained language models may suffer from inaccuracies in predicting boundaries， a boundary-aware approach for span-based extraction Machine Reading Comprehension （MRC） is proposed to mitigate this issue. Firstly， special characters were introduced to mark the question boundary during the question input stage， enhancing the semantic information of the question to improve boundary perception. Secondly， during the answer prediction stage， an answer boundary regressor was constructed to facilitate semantic interaction between the perceived question boundary and the output of the predicted answer boundary. Lastly， the biased predicted answer boundary was further adjusted based on the post-interaction semantic information to calibrate the predicted answers. Experimental results demonstrate that when compared to the SpanBERT （Span-based Bidirectional Encoder Representation from Transformers）， the proposed method improves the F1 value by 0.2 percentage points and the Exact Match （EM） value by 0.9 percentage points on the public dataset SQuAD （Stanford Question Answering Dataset）1.1， it achieved improvements of 0.7 percentage points in both F1 score and EM value on the HotpotQA （Hotpot Question Answering） dataset， and it improved the F1 score by 2.8 percentage points and the EM value by 3.3 percentage points on the NewsQA （News Question Answering） dataset. The effectiveness of this method is rooted in its capacity to enhance the model’s perception of question boundary information and to accomplish the calibration of predicted answer boundary. Consequently， it results in an enhancement of system accuracy in applications such as intelligent question answering and intelligent customer service when dealing with text data comprehension and analysis.

Aiming at the problem that the existing deep clustering methods can not efficiently divide event types without considering event information and its structural characteristics， a Deep Event Clustering method based on Event Representation and Contrastive Learning （DEC_ERCL） was proposed. Firstly， information recognition was utilized to identify structured event information from unstructured text， thus the impact of redundant information on event semantics was avoided. Secondly， the structural information of the event was integrated into the autoencoder to learn the low-dimensional dense event representation， which was used as the basis for downstream clustering. Finally， in order to effectively model the subtle differences between events， a contrast loss with multiple positive examples was added to the feature learning process. Experimental results on the datasets DuEE， FewFC， Military and ACE2005 show that the proposed method performs better than other deep clustering methods in accuracy and Normalized Mutual Information （NMI） evaluation indexes. Compared with the suboptimal method， the accuracy of DEC_ERCL is increased by 17.85%，9.26%，7.36% and 33.54%， respectively， indicating that DEC_ERCL has better event clustering effect.

To tackle the difficulty in semantic mining of entity relations and biased relation prediction in Relation Extraction （RE） tasks， a RE method based on Mask prompt and Gated Memory Network Calibration （MGMNC） was proposed. First， the latent semantics between entities within the Pre-trained Language Model （PLM） semantic space was learned through the utilization of masks in prompts. By constructing a mask attention weight matrix， the discrete masked semantic spaces were interconnected. Then， the gated calibration networks were used to integrate the masked representations containing entity and relation semantics into the global semantics of the sentence. Besides， these calibrated representations were served as prompts to adjust the relation information， and the final representation of the calibrated sentence was mapped to the corresponding relation class. Finally， the potential of PLM was fully exploited by the proposed approach through harnessing masks in prompts and combining them with the advantages of traditional fine-tuning methods. The experimental results highlight the effectiveness of the proposed method. On the SemEval （SemEval-2010 Task 8） dataset， the F1 score reached impressive 91.4%， outperforming the RELA （Relation Extraction with Label Augmentation） generative method by 1.0 percentage point. Additionally， the F1 scores on the SciERC （Entities， Relations， and Coreference for Scientific knowledge graph construction） and CLTC （Chinese Literature Text Corpus） datasets were remarkable， achieving 91.0% and 82.8% respectively. The effectiveness of the proposed method was evident as it consistently outperformed the comparative methods on all three datasets mentioned above. Furthermore， the proposed method achieved superior extraction performance compared to generative methods.

Aiming at the problem of incomplete semantic information of word vectors and the problem of word polysemy faced by text feature extraction， a BERT （Bidirectional Encoder Representation from Transformer） word vector-based Twice Attention mechanism weighting algorithm for Relation Extraction （TARE） was proposed. Firstly， in the word embedding stage， the self-attention dynamic encoding algorithm was used to capture the semantic information before and after the text for the current word vector by constructing Q， K and V matrices. Then， after the model output the sentence-level feature vector， the locator was used to extract the corresponding parameters of the fully connected layer to construct the relation attention matrix. Finally， the sentence level attention mechanism algorithm was used to add different attention scores to sentence-level feature vectors to improve the noise immunity of sentence-level features. The experimental results show that compared with Contrastive Instance Learning （CIL） algorithm for relation extraction， the F1 value is increased by 4.0 percentage points and the average value of Precision@100， Precision@200， and Precision@300 （P@M） is increased by 11.3 percentage points on the NYT-10m dataset. Compared with the Piecewise Convolutional Neural Network algorithm based on ATTention mechanism （PCNN-ATT）， the AUC （Area Under precision-recall Curve） value is increased by 4.8 percentage points and the P@M value is increased by 2.1 percentage points on the NYT-10d dataset. In various mainstream Distantly Supervised for Relation Extraction （DSRE） tasks， TARE effectively improves the model’s ability to learn data features.

With the application of artificial intelligence technology in the judicial field， charge prediction based on case description has become an important research content. It aims at predicting the charges according to the case description. The terms of case contents are professional， and the description is concise and rigorous. However， the existing methods often rely on text features， but ignore the difference of relevant elements and lack effective utilization of elements of action words in diverse cases. To solve the above problems， a multi-task learning model of charge prediction based on action words was proposed. Firstly， the spans of action words were generated by boundary identifier， and then the core contents of the case were extracted. Secondly， the subordinate charge was predicted by constructing the structure features of action words. Finally， identification of action words and charge prediction were uniformly modeled， which enhanced the generalization of the model by sharing parameters. A multi-task dataset with action word identification and charge prediction was constructed for model verification. The experimental results show that the proposed model achieves the F value of 83.27% for action word identification task， and the F value of 84.29% for charge prediction task； compared with BERT-CNN， the F value respectively increases by 0.57% and 2.61%， which verifies the advantage of the proposed model in identification of action words and charge prediction.

To address the problem that traditional PhotoVoltaic （PV） power prediction models are affected by random power fluctuation and tend to ignore important information， resulting in low prediction accuracy， ADDPG and ARDPG models were proposed by combining the attention mechanism with Deep Deterministic Policy Gradient （DDPG） and Recurrent Deterministic Policy Gradient （RDPG）， respectively， and a PV power prediction framework was proposed on this basis. Firstly， the original PV power data and meteorological data were normalized， and the PV power prediction problem was modeled as a Markov Decision Process （MDP）， where the historical power data and current meteorological data were used as the states of MDP. Then the attention mechanism was added to the Actor networks of DDPG and RDPG， giving different weights to different components of the state to highlight important and critical information， and learning critical information in the data through the interaction of Deep Reinforcement Learning （DRL） agents and historical data. Finally， the MDP problem was solved to obtain the optimal strategy and make accurate prediction. Experimental results on DKASC and Alice Springs PV system data show that ADDPG and ARDPG achieve the best results in Root Mean Square Error （RMSE）， Mean Absolute Error （MAE） and R². It can be seen that the proposed models can effectively improve the prediction accuracy of PV power， and can also be extended to other prediction fields such as grid prediction and wind power generation prediction.

Aiming at the problem of poor recognition of sentencing circumstances in adjudication documents caused by the lack of labeled data， low quality of labeling and existence of strong logicality in judicial field， a sentencing circumstance recognition model based on abductive learning named ABL-CON （ABductive Learning in CONfidence） was proposed. Firstly， combining with neural network and domain logic inference， through the semi-supervised method， a confidence learning method was used to characterize the confidence of circumstance recognition. Then， the illogical error circumstances generated by neural network of the unlabeled data were corrected， and the recognition model was retrained to improve the recognition accuracy. Experimental results on the self-constructed judicial dataset show that the ABL-CON model using 50% labeled data and 50% unlabeled data achieves 90.35% and 90.58% in Macro_F1 and Micro_F1， respectively， which is better than BERT （Bidirectional Encoder Representations from Transformers） and SS-ABL （Semi-Supervised ABductive Learning） under the same conditions， and also surpasses the BERT model using 100% labeled data. The ABL-CON model can effectively improve the logical rationality of labels as well as the recognition ability of labels by correcting illogical labels through logical abductive correctness.

Relation extraction aims to extract the semantic relationships between entities from the text. As the upper-level task of relation extraction， entity recognition will generate errors and spread them to relation extraction， resulting in cascading errors. Compared with entities， entity boundaries have small granularity and ambiguity， making them easier to recognize. Therefore， a relationship extraction method based on entity boundary combination was proposed to realize relation extraction by skipping the entity and combining the entity boundaries in pairs. Since the boundary performance is higher than the entity performance， the problem of error propagation was alleviated； in addition， the performance was further improved by adding the type features and location features of entities through the feature combination method， which reduced the impact caused by error propagation. Experimental results on ACE 2005 English dataset show that the proposed method outperforms the table-sequence encoders method by 8.61 percentage points on Macro average F1-score.

For addressing real root isolation problem of transcendental function polynomials， an interval isolation algorithm for exponential function polynomials named exRoot was proposed. In the algorithm， the real root isolation problem of non-polynomial real functions was transformed into sign determination problem of polynomial， then was solved. Firstly， the Taylor substitution method was used to construct the polynomial nested interval of the objective function. Then， the problem of finding the root of the exponential function was transformed into the problem of determining the positivity and negativity of the polynomial in the intervals. Finally， a comprehensive algorithm was given and applied to determine the reachability of rational eigenvalue linear system tentatively. The proposed algorithm was implemented in Maple efficiently and easily with readable output results. Different from HSOLVER and numerical calculation method fsolve， exRoot avoids discussing the existence of roots directly， and theoretically has termination and completeness. It can reach any precision and can avoid the systematic error brought by numerical solution when being applied into the optimization problem.

Empathy prediction from texts achieves little progress due to the lack of sufficient labeled data， while the related task of text sentiment polarity classification has a large number of labeled samples. Since there is a strong correlation between empathy prediction and polarity classification， a transfer learning?based text empathy prediction method was proposed. Transferable public features were learned from the sentiment polarity classification task to assist text empathy prediction task. Firstly， a dynamic weighted fusion of public and private features between two tasks was performed through an attention mechanism. Secondly， in order to eliminate domain differences in datasets between two tasks， an adversarial learning strategy was used to distinguish the domain?unique features from the domain?public features between two tasks. Finally， a Hinge?loss constraint strategy was proposed to make common features be generic for different target labels and private features be unique to different target labels. Experimental results on two benchmark datasets show that compared to the comparison transfer learning methods， the proposed method has higher Pearson Correlation Coefficient （PCC） and coefficient of determination （R²）， and has lower Mean?Square Error （MSE）， which fully demonstrates the effectiveness of the proposed method.

To solve the low efficiency of scheduling in injection molding workshop, an improved job-shop scheduling method was proposed based on clustering mold. The production time was reduced by merging jobs with the same tool list, and the energy consumption was reduced through small model injection machine preferred scheduling. The theoretical analysis and the experimental results show that the proposed mehtod can improve productivity and reduce power consumption more than 50%, making injection molding shop job scheduling be more efficient.

When Proximal Support Vector Machine (PSVM) deals with unbalanced samples, it will overfit the class with large samples and underestimate the misclassification error of the class with small samples, resulting in the decline of accuracy in overall samples. To solve this problem, a modified PSVM used for dealing with unbalanced samples was proposed. The new algorithm not only set different punishments for positive and negative samples, but also added a new parameter to the constraint, making the classification hyperplane more flexible. Firstly, the new algorithm trained the training set to obtain the optimal parameters, then the classification hyperplane was obtained by training the test set. Finally, the classification results was output. The experiments presented by 9 datasets in UCI database show that the new algorithm improves the classification accuracy of the samples, by 2.19 and 3.14 percentage points in the linear and nonlinear case respectively. The generalization ability of the algorithm is strengthened effectively.

Multi-chord curvature polynomial algorithm for corner detection was proposed based on Chord-to-Point Distance Accumulation (CPDA) technique and curvature product. Firstly, the edge map was extracted by Canny edge detector. Then, at each chord, a multi-chord curvature polynomial was used as the sum or multiplication of the contour curvature. The new method can not only effectively enhance curvature extreme peaks, but also prevent smoothing some corners. To reduce false or missing detection made by experiment threshold, local adaptive threshold was used to detect corners. According to the detection capability, localization accuracy and repeatability of corner number criteria, experiments were made to compare the proposed detector with several recent corner detectors. The experimental results demonstrate that the proposed detector has strong robustness, its detection accuracy increases by 14.5%, and its average repeatability increases by 12.6%.

Concerning the problem in which the fixed locomotion track captured from human movement can not be used in transformation between key postures for humanoid robot, a method of similar key posture transformation based on hierarchical Option for humanoid robot was proposed. The multi-level dendrogram of key postures was constructed and the difference of key postures was illustrated in respects of similar joint difference, moment total similar difference, period total similar difference. The hierarchical reinforcement Option learning was introduced, in which the sets of key postures and Option actions were constructed. SMDP-Q method tended to be the optimal Option function by the accumulative rewards of key posture difference and the transformations were realized. The experiments show the validity of the method.

Concerning the problems of lacking the mind states and task handling capability, low real-time caused by congested bandwidth and slow learning from negotiation history, a task negotiation model for multiple mobile-robots was proposed. Firstly, the basic moving state of robot was shown. Secondly, the states of mind (belief, goal, intention, knowledge update, etc.) and ability (cooperation, capability judgment, task allocation, etc.) based on π calculus for the negotiation of multiple mobile-robots were defined. Thirdly, the negotiation model of multiple mobile-robots was constructed, in which the negotiation period, negotiation task, utility estimation, negotiation allocation protocol, learning mechanism were studied. Finally, the validity of model was proved through experiments of robot soccer.

Traditional solutions deal with following-car interference and adjacent-lane interference in Electronic Toll Collection (ETC) system separately, and are of low efficiency and high cost. In order to solve this problem, a solution based on information fusion, which can deal with these two problems together, was proposed. This method used the vehicle information collected by an ETC system as known information and chose the features of the vehicle image to verify whether the vehicle information came from the car in question. Then D-S evidence theory was adopted to fuse the results of verification and make a final decision whether it was the right car to be charged. An improved method for D-S evidence theory was proposed to ensure its validity when the results of consistency verification conflicted with each other. The experimental results show that this method can highly reliably detect illegal vehicles and solve both following-car interference and adjacent lane interference problems in ETC system.

A new algorithm based on cyclic spectrum was proposed for classification of communication signals in multipath channel, which solved the problems of fewer identification types, difficults table feature parameters extraction and low recognition rate. Firstly, the features face and projective planes of cyclic spectrum, square cyclic spectrum and the fourth power cyclic spectrum were extracted. Secondly, correlation coefficients of features face and projective planes were used as the characteristic parameters. At last, the suitable decision threshold was chosen and seven signals of BPSK, QPSK, 2FSK, 4FSK, MSK, 16QAM and OFDM were identified automatically. The experimental results show that the characteristic parameters have great ability for multipath interference and high recognition rate is acquired at last. When the Signal-to-Noise Ratio (SNR) is higher than 2dB, its overall recognition rate is up to 94%. Compared with the existing algorithms, the simulation results prove that the algorithm is superior.

The coverage problem is a basic problem in the wireless sensor networks, which indicates the Quality of Service (QoS) of sensing by wireless sensor networks. A lot cover blind areas and cover redundancies will be produced, when the nodes are deployed initially in the networks. A hybrid algorithm was proposed to deploy the heterogeneous network nodes reasonably to improve the coverage ratio and reduce the cost of the nodes,which introduced the ε-target constraint method based on Particle Swarm Optimization (PSO) and Fish Swarm Algorithm (FSA). The swarm hybrid algorithm firstly set up the concept of individual center, to quickly search the best solution domain of the individuals' locations, introducing the idea of the cluster behavior and tracing cauda behavior into the PSO, and then used the PSO to find the optimized speed and optimized location of the individuals. The simulation results show that the swarm hybrid algorithm is better than the standard PSO and the standard FSA in pursuing the balance and optimization between the coverage ratio and the cost of the networks.

To segment moving crowd with different dynamics in complex video surveillance scenes, this paper proposed a crowd motion segmentation algorithm which was based on video particle flow and Finite Time Lyapunov Exponent (FTLE) field. Firstly, video particle flow was used to represent the long-range particle motion estimation. To optimize these particles trajectories, an energy function containing point-based appearance matching and distortion between the particles was minimized. Then the spatial gradient of the particle flow map was solved and the FTLE field was constructed. Finally, the Lagrangian Coherent Structure (LCS) in the FTLE field was used to divide flow into regions of qualitatively different dynamics. The experimental results show that the proposed algorithm can effectively segment crowd flow with different dynamics in complex video surveillance scenes, and it has strong robustness.

When using recognition algorithm of high-order cumulants to classify and recognize digital modulation signals, the calculation of six-order and six-order above cumulants are too complex and the signals of 8PSK and Multiple Frequency Shift Keying (MFSK) have the same cumulants, so it is impossible to recognize directly. To solve this problem, a new classification algorithm was proposed in this paper, which made wavelet transform on MFSK and 8PSK at first, and then used four-order cumulants recognition. The simulations show that the characteristic parameters could restrain Gaussian white noise efficiently and simply, and classify and recognize 2ASK/BPSK, 4ASK, 2FSK, 4FSK, QPSK, 8PSK and 16QAM successfully. When SNR (Signal-to-Noise Ratio) is 3dB, the recognition rate reaches as high as 96%. Compared with the existing algorithms, the superiority of the algorithm is proved.

In the super-peer network, the super-peer topology structure and its mechanism of dynamic maintenance and search routing are important factors affecting network performance and search efficiency. In this paper, a new structure named k-Perfect Difference Graph (PDG) was proposed by analyzing the characteristics and the deficiencies of PDG, new Super-peer Network based on k-PDG (KPDGN) was constructed, and then the mechanism of dynamic maintenance and search routing was presented in KPDGN. The analysis and simulation results show that compared with current supper-peer topology, KPDGN has good performance with constant degree and fixed adjacent nodes, which reduces the bandwidth consumption during searching and the cost of topology construction and maintenance.

Concerning the inequality of cluster-head distribution and node energy consumption in Wireless Sensor Network (WSN) cluster routing protocol, a node-energy load-balance clustering algorithm was proposed. Group according to the node energy, dynamically adjust the group number to the node energy reduction, conduct cluster-head election in the group according to the energy focus, and further balance the node energy consumption using cluster-head rotate and multi-hop routing between clusters. The simulation results show that this protocol effectively balances the energy consumption among network nodes and achieves an obvious improvement in network stable period.

A multi-source overlay multicast model was built to solve the problem of Quality of Service (QoS) multicast in overlay network. In this model, each node maintained the local state information instead of one tree per source, and the model delivered multicast message in a way similar to flooding. By controlling, the message transmission path formed a tree structure. The multicast tree can dynamically adjust and adapt to a different source node initiated multicast applications, which meets the requirements of QoS. The experimental results show that multicast coverage rate of the model is high, and the cost of adjusting the multicast tree can be the lowest by controlling the number of sub-node of modes.

RP2P algorithm combines arbitrary neighbor selection, typically used only in unstructured P2P networks, with a Distributed Hash Table (DHT) ring. It is the first of its kind to resolve requests in d hops with a chosen probability of 1－c. However, the capacities of the hosts participating in the network, such as bandwidth, memory, CPU, are very different, which will affect the efficiency of the whole system. On the other hand, the shock caused by some of the nodes in the network frequent joining in/departing from the system is also one of the factors affecting the performance. This paper analyzed the capacities of the nodes and proposed an efficient RP2P network based on hierarchical dividing. It improves the efficiency of the system, and solves the problem of system shocks.

Histogram approach often leads to many false detections and missing detections in shot boundary detection, while edge feature-based approach suffers high computation cost. An improved detection approach that took advantage of the two approaches was proposed, namely, first the shot boundary was detected using histograms, and then based on previous results the second detection was applied with edge feature and various step length. The experimental results show that the new method increases recall and precision without increasing computation cost.

With diversification analyses of lost customers in applications, a multi-mode classification algorithm was presented. It applied clustering algorithm to segment the lost customers into many subgroups, then used classification algorithm to build the classifying models on each subgroup. Meanwhile, it filtered the low-precision class models to ensure the precision improvement of forecast lists. Compared with Logistic, decision tree and neural network, the experimental results show that the new classification algorithm has better performance.

Having studied the advantages and disadvantages of some traditional searching methods in structured and unstructured P2P networks, an assisted interest-based searching method was proposed to enhance the search in unstructured P2P overlay networks by registering interests of nodes in structured P2P overlay networks. Experimental results show that this method achieves good performance in success rate and search latency.