Search results

This paper proposes a self-supervised monocular depth estimation algorithm under multiple constraints, which can generate the corresponding depth map end-to-end based on RGB images. On this basis, based on the traditional visual simultaneous localisation and mapping (VSLAM) framework, a dynamic object detection framework based on deep learning is introduced, and dynamic objects in the scene are culled during mapping.

Typical SLAM algorithms or data sets assume a static environment and do not consider the potential consequences of accidentally adding dynamic objects to a 3D map. This shortcoming limits the applicability of VSLAM in many practical cases, such as long-term mapping. In light of the aforementioned considerations, this paper presents a self-supervised monocular depth estimation algorithm based on deep learning. Furthermore, this paper introduces the YOLOv5 dynamic detection framework into the traditional ORBSLAM2 algorithm for the purpose of removing dynamic objects.

Compared with Dyna-SLAM, the algorithm proposed in this paper reduces the error by about 13%, and compared with ORB-SLAM2 by about 54.9%. In addition, the algorithm in this paper can process a single frame of image at a speed of 15–20 FPS on GeForce RTX 2080s, far exceeding Dyna-SLAM in real-time performance.

This paper proposes a VSLAM algorithm that can be applied to dynamic environments. The algorithm consists of a self-supervised monocular depth estimation part under multiple constraints and the introduction of a dynamic object detection framework based on YOLOv5.

The purpose of this paper is to propose an improved charge simulation method (CSM) based on the real‐coded genetic algorithm (GA), in which the tedious testing and adjusting work in the traditional CSM is avoided. In addition, less simulation charges are used in the simulation model for the same object compared with the traditional CSM.

In the improved CSM, the information of testing points are combined with the matching points, hence the size and locations of simulation charges can be only computed by the information of matching points, and the testing calculation in the traditional CSM is avoided. According to the Maxwell equations, an overdetermined equation is reformulated, and an improved GA is used to solve it. The process to create the initial population is improved, boundary condition information of part matching points and human experience are added into the initial population, which enhanced the convergence rate of the search calculation.

Combining the information of testing points into that of the matching points in the improved CSM, the tedious testing calculation can be avoided, and the GA is effective to solve the overdetermined equation for the improved CSM. By importing the information of part matching points and human experience, the search process can be accelerated and the convergence of the method is ensured. Comparison between the efficiency of traditional CSM and the improved CSM is presented, less simulation charges are used compared with the traditional CSM for the same object.

The improved CSM cannot adjust the count of simulation charges in the equivalent model automatically.

Calculations of a sphere‐plane electrode model and a 64 kV insulator string are carried out to verify the validity of the improved CSM. In addition, it can also be used to compute the power frequency electric field of HV transmission apparatuses.

By improving the process of the traditional CSM and importing the GA, an improved CSM for the calculation of power frequency electric fields produced by high‐voltage apparatuses is presented in the paper, which can avoid the tedious testing and adjusting work in the traditional CSM.

Human-induced marine environmental noise, such as commercial shipping and seismic exploration, is concentrated in the low-frequency range. Meanwhile, low-frequency sound signals can achieve long-distance propagation in water. To meet the requirements of long-distance underwater detection and communication, this paper aims to propose an micro-electro-mechanical system (MEMS) flexible conformal hydrophone for low-frequency underwater acoustic signals. The substrate of the proposed hydrophone is polyimide, with silicon as the piezoresistive unit.

This paper proposes a MEMS heterojunction integration process for preparing flexible conformal hydrophones. In addition, sensors prepared based on this process are non-contact flexible sensors that can detect weak signals or small deformations.

The experimental results indicate that making devices with this process cannot only achieve heterogeneous integration of silicon film, metal wire and polyimide, but also allow for customized positions of the silicon film as needed. The success rate of silicon film transfer printing is over 95%. When a stress of 1 Pa is applied on the x-axis or y-axis, the maximum stress on Si as a pie-zoresistive material is above, and the average stress on the Si film is around.

The flexible conformal vector hydrophone prepared by heterogeneous integration technology provides ideas for underwater acoustic communication and signal acquisition of biomimetic flexible robotic fish.

The paper's goal is to examine and illustrate the useful uses of submodeling in finite element modeling for topology optimization and stress analysis. The goal of the study is to demonstrate how submodeling – more especially, a 1D approach – can reliably and effectively produce ideal solutions for challenging structural issues. The paper aims to demonstrate the usefulness of submodeling in obtaining converged solutions for stress analysis and optimized geometry for improved fatigue life by studying a cantilever beam case and using beam formulations. In order to guarantee the precision and dependability of the optimization process, the developed approach will also be validated through experimental testing, such as 3-point bending tests and 3D printing. Using 3D finite element models, the 1D submodeling approach is further validated in the final step, showing a strong correlation with experimental data for deflection calculations.

The authors conducted a literature review to understand the existing research on submodeling and its practical applications in finite element modeling. They selected a cantilever beam case as a test subject to demonstrate stress analysis and topology optimization through submodeling. They developed a 1D submodeling approach to streamline the optimization process and ensure result validity. The authors utilized beam formulations to optimize and validate the outcomes of the submodeling approach. They 3D-printed the optimized models and subjected them to a 3-point bending test to confirm the accuracy of the developed approach. They employed 3D finite element models for submodeling to validate the 1D approach, focusing on specific finite elements for deflection calculations and analyzed the results to demonstrate a strong correlation between the theoretical models and experimental data, showcasing the effectiveness of the submodeling methodology in achieving optimal solutions efficiently and accurately.

The findings of the paper are as follows: 1. The use of submodeling, specifically a 1D submodeling approach, proved to be effective in achieving optimal solutions more efficiently and accurately in finite element modeling. 2. The study conducted on a cantilever beam case demonstrated successful stress analysis and topology optimization through submodeling, resulting in optimized geometry for enhanced fatigue life. 3. Beam formulations were utilized to optimize and validate the outcomes of the submodeling approach, leading to the successful 3D printing and testing of the optimized models through a 3-point bending test. 4. Experimental results confirmed the accuracy and validity of the developed submodeling approach in streamlining the optimization process. 5. The use of 3D finite element models for submodeling further validated the 1D approach, with specific finite elements showing a strong correlation with experimental data in deflection calculations. Overall, the findings highlight the effectiveness of submodeling techniques in achieving optimal solutions and validating results in finite element modeling, stress analysis and optimization processes.

The originality and value of the paper lie in its innovative approach to utilizing submodeling techniques in finite element modeling for structural analysis and optimization. By focusing on the reduction of finite element models and the creation of smaller, more manageable models through submodeling, the paper offers designers a more efficient and accurate way to achieve optimal solutions for complex problems. The study's use of a cantilever beam case to demonstrate stress analysis and topology optimization showcases the practical applications of submodeling in real-world scenarios. The development of a 1D submodeling approach, along with the utilization of beam formulations and 3D printing for experimental validation, adds a novel dimension to the research. Furthermore, the paper's integration of 1D and 3D submodeling techniques for deflection calculations and validation highlights the thoroughness and rigor of the study. The strong correlation between the finite element models and experimental data underscores the reliability and accuracy of the developed approach. Overall, the originality and value of this paper lie in its comprehensive exploration of submodeling techniques, its practical applications in structural analysis and optimization and its successful validation through experimental testing.

Community question answering (CQA) platforms play a significant role in knowledge dissemination and information retrieval. Expert recommendation can assist users by helping them find valuable answers efficiently. Existing works mainly use content and user behavioural features for expert recommendation, and fail to effectively leverage the correlation across multi-dimensional features.

To address the above issue, this work proposes a multi-dimensional feature fusion-based method for expert recommendation, aiming to integrate features of question–answerer pairs from three dimensions, including network features, content features and user behaviour features. Specifically, network features are extracted by first learning user and tag representations using network representation learning methods and then calculating questioner–answerer similarities and answerer–tag similarities. Secondly, content features are extracted from textual contents of questions and answerer generated contents using text representation models. Thirdly, user behaviour features are extracted from user actions observed in CQA platforms, such as following and likes. Finally, given a question–answerer pair, the three dimensional features are fused and used to predict the probability of the candidate expert answering the given question.

The proposed method is evaluated on a data set collected from a publicly available CQA platform. Results show that the proposed method is effective compared with baseline methods. Ablation study shows that network features is the most important dimensional features among all three dimensional features.

This work identifies three dimensional features for expert recommendation in CQA platforms and conducts a comprehensive investigation into the importance of features for the performance of expert recommendation. The results suggest that network features are the most important features among three-dimensional features, which indicates that the performance of expert recommendation in CQA platforms is likely to get improved by further mining network features using advanced techniques, such as graph neural networks. One broader implication is that it is always important to include multi-dimensional features for expert recommendation and conduct systematic investigation to identify the most important features for finding directions for improvement.

This work proposes three-dimensional features given that existing works mostly focus on one or two-dimensional features and demonstrate the effectiveness of the newly proposed features.

The purpose of this study is to investigate whether the corporate environmental, social and governance (ESG) performance of enterprise is influenced by the enterprise digital transformation. In addition, this study explains how enterprise digital transformation affects ESG performance.

The sample covers 4,646 nonfinancial companies listed on China’s A-share market from 2009 to 2021. The study adopts the fixed-effects multiple linear regression to perform the data analysis.

The study finds that enterprise digital transformation has a significant inverted U-shaped impact on ESG performance. Moderate digital transformation can improve enterprise ESG performance, whereas excessive digital transformation will bring new organizational conflicts and increase enterprise costs, which is detrimental to ESG performance. This inverted U-shaped effect is more pronounced in industrial cities, manufacturing industries and enterprises with less financing constraints and executives with financial backgrounds. Enterprise digital transformation mainly affects ESG performance by affecting the level of internal information communication and disclosure, the level of internal control and the principal-agent cost.

The government should take multiple measures to encourage enterprises to choose appropriate digital transformation based on their own production behaviors and development strategies, encourage them to innovate and upgrade their organizational management and development models in conjunction with digital transformation and guide them to use digital technology to improve ESG performance.

This study shows that irrational digital transformation cannot effectively improve the ESG performance of enterprises and promote the sustainable development of the country. Enterprises should carry out reasonable digital transformation according to their own development needs and finally improve the green and sustainable development ability of enterprises and promote the sustainable development of society.

This study examines the relationship between enterprise digital transformation and ESG performance. Different from the linear relationship between the two in previous major studies, this study proves the inverse U-shaped relationship between enterprise digital transformation and ESG performance through mathematical theoretical model derivation and empirical test. This study also explores in detail how corporate digital transformation affects ESG performance, as well as discusses heterogeneity at the city, industry and firm levels. It is proposed that enterprises should take into account their own characteristics and carry out reasonable digital transformation according to their development needs.