AUTOMATIC CONTROL AND COMPUTER SCIENCES最新文献

This paper briefly described the end-to-end machine translation algorithm based on long short-term memory (LSTM), and introduced part-of-speech auxiliary information into machine translation. The translation algorithm performance under different hidden layer numbers and node numbers and the improved machine translation algorithm designed in this article were tested in simulation experiments. Part-of-speech auxiliary information ablation experiments were conducted. It was found that the translation algorithm achieved optimal performance when there were two hidden layers in the LSTM of the encoder and decoder and 256 hidden layer nodes in each layer. The proposed algorithm had the highest bilingual evaluation understudy (BLEU) score, followed by the LSTM-based algorithm, and the recurrent neural network-based algorithm had the lowest score. The performance of the proposed machine translation algorithm decreased after excluding any part-of-speech, and the exclusion of the verb had the largest impact.

With the rapid progress of information technology, information has evolved from its original textual form to a rich and diverse form of internet information. It is fully reflected in smart home applications in the intuitive form of images. However, image search in smart homes is a real-time process. Existing systems cannot meet the requirement for users to receive responses in a short period of time. Therefore, in response to the long system response time, long image processing time, and low retrieval accuracy, an image search system based on parallel computing of the central processor is proposed for application in smart home design. Firstly, the image area is segmented and preprocessed. Secondly, a multilevel edge detection algorithm for parallel computing between the central processing unit and graphics processor is designed. Finally, an image search system with parallel computing between the central processing unit and graphics processor is established. The research results indicate that the increase in central processing unit running time is the most significant, increasing from 0.18 to 1.28, with a growth rate of 1.1. The growth rate of graphics processor runtime is relatively small, increasing from 0.05 to 0.17, with a growth rate of 0.12. As the image size increases, the overall computational complexity also increases. The parallel acceleration performance of graphics processors and central processing units is gradually becoming significant.

In this paper we delve into the conceptual underpinnings of fuzzy logic and its applicability to forecasting within the context of nonscheduled passenger air transportation. We review relevant literature, highlighting the limitations of traditional forecasting techniques and the rationale for adopting a fuzzy approach. Additionally, we outline the methodology employed in developing the proposed fuzzy forecasting model, emphasizing its adaptability to evolving operational conditions and its potential to enhance decision-making processes within the aviation industry. In the conducted research, a new method of building a forecasting model using a fuzzy approach was proposed for the time series of nonscheduled passenger air transportation with intraseries multiplicative changes. The method is based on the use of membership functions in the calculation of forecast values based on statistical indicators of intrarow changes. In regular air transportation, the intraseries changes of statistical indicators of time series are stable. On charter flights, these changes are unstable. This is due to the strong random effects of external factors (a sudden increase in demand for flights, economic changes, etc.) on the formation of charter flights. For this reason, the application of models based on trend changes does not give good enough results when building forecast models in charter air transportation. Therefore, to solve the problem, we propose to build the forecasting model of nonscheduled passenger air transportation using a fuzzy approach. The researched method was checked based on the actual data of the time series of charter flights. The obtained results were compared with classical forecasting models (ARIMA, Fine, Medium, and Coarse SVM), and it was noted that the results were obtained within acceptable limits.

The heterogeneity and feasibility of test scenarios are crucial factors in controlled experiments examining the performance of autonomous vehicles. The challenges in constructing a test site or outdoor laboratory for such experiments lie in efficiently and effectively planning and testing the critical test scenarios. This paper proposes a heuristics path planning method based on the rapidly exploring random-tree algorithm for solving the challenge mentioned above. To demonstrate this new path planning method, a case study is conducted at the outdoor laboratory of Suzhou Automotive Research Institute, Tsinghua University, China. The results show that the new path planning method not only allows more test scenarios to be implemented but also completes all of the necessary experiments within the shortest total experimental time or mileage. Hence, the new path planning method saves costs, increases the efficiencies of controlled experiments, and accelerates the scenario testing of autonomous vehicles in closed-form test sites.

The development of graphical user interfaces has made significant progress in the past few decades, playing an important role in computer user experience and human-computer interaction. However, at present, there is a lack of professional experienced workers in graphical user interfaces, and the art design of graphical user interfaces has low attention in real life. Therefore, this research introduces reinforcement learning algorithm, combines it with deep network, and realizes automation and intelligence in the generation of art design oriented graphical user interface and graphical user interface wireframe. The test results indicate that the graphical user interface method proposed in this study has average values of 0.075 and 0.869 for the Fréchet inception distance and one nearest neighbor accuracy in the category subset, and 0.070 and 0.823 for the development company subset. The comprehensive average scores for the three indicators of aesthetics, color coordination, and structure in manual evaluation are 3.11, 3.30, and 3.21, respectively. The research proposes a wireframe generation method with average values of Fréchet inception distance and one nearest neighbor accuracy of 0.082 and 0.911, respectively. The average value of position deviation index is 1.018, the average score of manual evaluation is 3.32, and the average values of structural similarity and spatial Euclidean distance are 0.363 and 3.683. The experimental results indicate that the method designed in this study generates a graphical user interface with higher quality than traditional common methods, and is more aesthetically pleasing, in line with popular art aesthetics.

Finding a way to solve the trilemma, which requires striking a balance between scalability, security, and decentralization, is a persistent problem in the field of blockchain technology. In order to overcome this trilemma, this study presents a novel blockchain architecture that combines cutting-edge cryptography techniques, creative security protocols, and flexible decentralization mechanisms. Our framework is a new standard for secure, scalable, and decentralized blockchain ecosystems. It utilizes well-known techniques like zero knowledge proof (zk-SNARK), Schnorr VRF, elliptic curve cryptography (ECC), and in addition to innovative approaches for anomaly detection, incentive alignment, and stake distribution. The suggested system outperforms elite consensuses by obtaining 1600+ TPS, guaranteeing strong security against all known blockchain attacks without sacrificing scalability, and obtaining a strong decentralization score of 7.181, which, when compared to other blockchain systems in benchmark analysis, shows strong decentralization.

Remote data monitoring of hydraulic pump stations holds significant value for fault diagnosis and prediction. This paper presents a monitoring system for hydraulic pump stations utilizing an Internet of Things (IoT) module. The system employs the ESP8266 IoT data module and utilizes the STM32F767VGT6 microcontroller as the primary controller, facilitating data transmission and exchange through wireless communication. The monitoring client system is developed on the Alibaba Cloud IoT Studio platform, enabling connection to the cloud server for data monitoring and command transmission. Experimental validation of the system demonstrates stable operation, remote data collection, and cloud storage capabilities. Additionally, the testing errors for pressure, oil temperature, oil level, and flow rate remain below 0.6%, indicating high measurement accuracy. This design offers valuable options for the fault diagnosis and prediction of hydraulic pump stations.

Video target tracking has gained a lot of interest and applications due to the quick development of computer vision and artificial intelligence. Adaptive modified target tracking approach based on target prediction algorithm and deep reinforcement learning is researched to realize exact positioning of the occluded target and to increase the efficiency, precision, and accuracy of real-time tracking of video targets. And combined with secondary correlation, a multitarget tracking algorithm is proposed to realize target tracking accuracy. The validation experiments are conducted in this research, and the findings indicate that the target tracking effect is at its greatest when the weight adjustment coefficient (p = 0.061) is attained, along with the peak area ratio and similarity of the correlation filtering response reaching their ideal advantage. The target frame only needs to move less than 5 movements in most of the images to successfully capture the target. It is found that the tracking accuracy of the proposed research method has comparable tracking accuracy with the MDNet with optimal performance, while the processing efficiency is improved by 80%, which is an accurate and efficient target tracking method. It is useful as a reference for target recognition in video and has some relevance for target localisation research in subsequent tracking systems.

For addressing the issues of ineffective mining and memory overflow when dealing with high-dimensional and large-scale datasets with traditional comparative pattern mining, and to further lift the limitation of a single machine’s own hardware, the study proposes a parallel comparative pattern mining algorithm based on Spark cluster environment. By constructing an extended data collection project tree, introducing an optimised decision tree for mining, and improving the related load balancing strategy, effective mining of large-scale and high-dimensional datasets is achieved. Experiments show that the algorithm proposed in the study has a maximum value of 1883 and a minimum value of 1549 for the number of contrasting patterns mined in the small-scale and low-dimensional Mushroom dataset, which is slightly higher than the mining method of strong jump revealed patterns with good classification performance. In the large-scale and high-dimensional dataset US census1990, the overall running time of the algorithm of the study is low compared to the cryptogrowth algorithm (T_max 43.2 min, T_min 18.4 min), and finally the failure request rate of the algorithm itself and the improved and weighted polling algorithms are ompared separately, and the results show that the improved algorithm takes the lowest time of 4%. The experiment showcases that the classification effect of the studied algorithm is good, the load balancing strategy of the improved algorithm is effective, and the overall performance of the algorithm is good.

Based on its unique memory characteristics, nonlinear characteristics, and nanoscale structure, fractional-order memristor has broad application value in many fields. So it has become a research hotspot in recent years. Firstly, this paper analyzed the waveforms of conductivity and voltage-current characteristic curves of fractional-order memristors with order changed in the time domain. Then, based on the particle swarm optimization algorithm combined with the Matlab function fminsearch(), the transfer function model of the fractional-order memristor was identified in the frequency domain. The identification effect was demonstrated to be good by fitting the curves and the value of the objective function. Finally, a fractional-order optimum (P{{I}^{lambda }}{{D}^{mu }}) controller was designed for the fractional-order memristor model. By controlling indicators, it is demonstrated that the control effect is much better than integer-order PID control.