Generation Algorithm Research Articles

AerialIRGAN: unpaired aerial visible-to-infrared image translation with dual-encoder structure

Due to the high cost of equipment and the constraints of shooting conditions, obtaining aerial infrared images of specific targets is very challenging. Most methods using Generative Adversarial Networks for translating visible images to infrared greatly depend on registered data and struggle to handle the diversity and complexity of scenes in aerial infrared targets. This paper proposes a one side end-to-end unpaired aerial visible-to-infrared image translation algorithm, termed AerialIRGAN. AerialIRGAN introduces a dual-encoder structure, where one encoder is designed based on the Segment Anything Model to extract deep semantic features from visible images, and the other encoder is designed based on UniRepLKNet to capture small-scale patterns and sparse patterns from visible images. Subsequently, AerialIRGAN constructs a bridging module to deeply integrate the features of both encoders and their corresponding decoders. Finally, AerialIRGAN proposes a structural appearance consistency loss to guide the synthetic infrared images to maintain the structure of the source image while possessing distinct infrared characteristics. The experimental results show that compared to the existing typical infrared image generation algorithms, the proposed method can generate higher-quality infrared images and achieve better performance in both subjective visual description and objective metric evaluation.

Influence of Fiber Volume Fraction on the Predictability of UD FRP Ply Behavior: A Validated Micromechanical Virtual Testing Approach.

Enhancing the understanding of the behavior, optimizing the design, and improving the predictability and reliability of manufactured unidirectional (UD) FRP plies, which serve as primary building blocks for structural FRP laminates and components, are crucial to achieving a safe and cost-effective design. This research investigated the influence of fiber volume fraction (vf) on the predictability and reliability of the homogenized elastic properties and damage initiation strengths of two different types of UD FRP plies using validated micromechanical virtual testing for representative volume element (RVE) models. Several sources of uncertainties were included in the RVE models. This study also proposed a modified algorithm for microstructure generation and explored the effect of vf on the optimal sizes of the RVE in terms of fiber number. Virtual tests were systematically conducted using full factorial DOE coupled with Monte Carlo simulation. The modified algorithm demonstrated exceptional performance in terms of convergence speed and jamming limit, significantly reducing the time required to generate microstructures. The developed RVE models accurately predicted failure modes, loci, homogenized elastic properties, and damage initiation strengths with a mean error of less than 5%. Also, it was found that increasing vf led to a concurrent increase in the optimal size of the RVE. While it was found that the vf had a direct influence on homogenized elastic properties and damage initiation strengths, it did not significantly affect the reliability and predictability of these properties, as indicated by low correlation coefficients and fluctuations in the coefficient of variation of normalized properties.

Targeted iterative assembly and the midpoint method for improved convergence in point-by-point algorithms for freeform generation

We present the target iterative assembly (TIA), an algorithm to generate freeform lenses for non-imaging systems from prescribed input rays and a distribution of targets, whose coordinates are functions of the positional and angular variables of the source distribution, granting the optical design unlimited customization and control over aberrations, whereas the angular iterative assembly (AIA) presented in a previous work was restricted to choosing output angles. The formalism employed seeks to generalize a class of existing point-by-point iterative construction methods. We also elaborate variations of the midpoint method, a predictor-corrector approach to enhance convergence of the solutions with fewer rays and analyze the sensibility of their error profiles to changes in the initial parameters. For N R rays, we obtained surface average errors and computing times proportional to (10 N R )−2 and 10−5 N R seconds, respectively, allowing for quick solutions within the precision limits of fabrication. We also discuss two algorithms for the generation of multiple successive surfaces, one which leads to apodization, and set the base for further developments, such as a 3D version to properly generate freeforms. The Python code with fully customizable system parameters is available.

Global supply chain flow planning for Chinese manufacturing under the BRI: An SFG-DRO method

The Belt and Road Initiative (BRI) has fostered the free flow of commodities and optimal distribution of resources for the global manufacturing supply chain (GMSC), meanwhile confronted with risks and uncertainties. This paper proposes the SFG-DRO method, which can well solve the flow distribution problem of GMSC under the BRI with uncertainties. The proposed method combines trend prediction from the stochastic frontier gravity (SFG) model with 1-Wasserstein fuzzy set-based distributionally robust optimization (DRO). Additionally, this paper integrates the principles of the column and constraint generation (C&CG) algorithm with the particle swarm optimization (PSO) algorithm to handle the above two-stage nonlinear problems efficiently. The results show that the SFG-DRO distribution method significantly reduces transportation costs by 25% and enhances security. By examining flow distribution across three key BRI channels, this paper identifies underexploited markets along the southward channel and substantial market potential with security risks along the northeast sea-land channel, proposing relevant improvement suggestions.

The impact and application of artificial intelligence technology on mental health counseling services for college students

With the intensification of social pressure and the enhancement of mental health awareness, the mental health issues of college students have become increasingly prominent, attracting social attention. Mental health counseling services, as an important way to alleviate students’ psychological stress, are facing the dual challenges of a shortage of professionals and growing service demands. In recent years, the application of artificial intelligence (AI) technology in the field of mental health has gradually risen, and its advantages in data analysis, pattern recognition, and natural language processing provide new solutions for mental health counseling services. However, existing research still faces problems such as insufficient understanding and limited emotional interaction capabilities in practical applications. This paper delves into the application of AI technology in mental health counseling services for college students and innovates and improves upon the deficiencies in existing research. The study focuses on two main areas: First, word vector generation technologies based on statistics and language models are used according to different application scenarios, and their effectiveness in the analysis of mental health counseling texts is compared. Second, an improved Seq2Seq model is proposed to enhance the emotional understanding and interaction capabilities of emotional dialogue generation algorithms in mental health counseling. This study not only provides technological support for college mental health counseling services but also opens new research directions and perspectives for the application of AI in the field of mental health.

Proactive Prevention Strategies for Grid Risk Scenarios Based on Pumped Storage Participation

Given the increasing importance of cascading trip prevention, this study presents the pumped storage characteristics to cooperate with thermal units and further ensure the safety of grid operation. The proposed model comprises two phases: the first phase is the day-ahead decision-making phase, which synergistically optimizes thermal units and pumped storage to improve the flexibility of grid operation; the second phase is the intra-day adjustment phase considering the uncertainty of the wind-photovoltaic and transmission line failure outage to enhance the robustness. Furthermore, this model employs a column and constraint generation algorithm to solve the two-phase, three-layer optimization problem, thereby minimizing the adjustment cost under the riskiest scenarios. The experimental results show that adding pumped storage reduces the total system operating cost by 3.99%, the renewable energy abandonment rate to 0.65%, and the thermal power unit output standard deviation to 74.13 MW.

Agent-based digital traffic model generation for regions facing data scarcity using aggregated cellphone data: a case study for Brussels

ABSTRACT The adoption of agent-based models (ABMs) in transport research has highlighted their potential in digitally simulating travel behaviour from a bottom-up perspective. However, generating ABMs remains a complex multi-step process, often suffering from the lack of model representation or reliance on proprietary data, thereby limiting ABMs' transferability, especially in regions facing data scarcity. This research presents a systematic ABM generation algorithm for Brussels, Belgium, that generates high-quality ABMs more representative of real-life travel patterns, including incorporating external travel demand and allowing multiple destinations for the same activity types in activity chaining. Additionally, our algorithm's input data only contains social-spatial data and aggregated cellphone matrices. Due to the ubiquitous nature of such data worldwide, our algorithm demonstrates significant transferability to other regions for ABM generation while preserving individual privacy. We demonstrate the input data and the proposed algorithm step by step. The ABM generated using MATSim is validated against real-world data regarding mode share, distance share, travel time distribution and traffic counts. Overall, this research would serve as valuable guidance for ABM modellers in data collection and specific modelling setups, lowering the entry barrier of ABM research towards more efficient, representative and reproducible ABMs.

Species selection for automatic chemical kinetic mechanism generation

AbstractMany important chemical kinetic systems require detailed chemical kinetic models to resolve. These detailed kinetic models can involve thousands of species and hundreds of thousands of chemical reactions, making them difficult to construct by hand. Modern automatic mechanism generation algorithms can mostly be divided into two classes: rule and rate based. Rule‐based generators choose species based on user defined constraints on species and reaction classes. Rate‐based generators generate a much larger set of potentially important species and reactions and then choose which ones to add based on running simulations of species and reactions deemed important and calculating the flux to potentially important species. In principle, the latter is preferable, as it requires the user to make far fewer assumptions about what is important in the system. However, while the effectiveness of the rate‐based approach has been demonstrated in a wide variety of systems, it has also been demonstrated to have difficulty picking up important low‐flux chemistries. Here we present a discussion of the challenges associated with rate‐based mechanism generation and new algorithms that are able to efficiently mitigate these challenges improving species selection during mechanism generation in a set of case studies.

Chemical-Inspired Material Generation Algorithm (MGA) of Single- and Double-Diode Model Parameter Determination for Multi-Crystalline Silicon Solar Cells

The optimization of solar photovoltaic (PV) cells and modules is crucial for enhancing solar energy conversion efficiency, a significant barrier to the widespread adoption of solar energy. Accurate modeling and estimation of PV parameters are essential for the optimal design, control, and simulation of PV systems. Traditional optimization methods often suffer from limitations such as entrapment in local optima when addressing this complex problem. This study introduces the Material Generation Algorithm (MGA), inspired by the principles of material chemistry, to estimate PV parameters effectively. The MGA simulates the creation and stabilization of chemical compounds to explore and optimize the parameter space. The algorithm mimics the formation of ionic and covalent bonds to generate new candidate solutions and assesses their stability to ensure convergence to optimal parameters. The MGA is applied to estimate parameters for two different PV modules, RTC France and Kyocera KC200GT, considering their manufacturing technologies and solar cell models. The significant nature of the MGA in comparison to other algorithms is further demonstrated by experimental and statistical findings. A comparative analysis of the results indicates that the MGA outperforms the other optimization strategies that previous researchers have examined for parameter estimation of solar PV systems in terms of both effectiveness and robustness. Moreover, simulation results demonstrate that MGA enhances the electrical properties of PV systems by accurately identifying PV parameters under varying operating conditions of temperature and irradiance. In comparison to other reported methods, considering the Kyocera KC200GT module, the MGA consistently performs better in decreasing RMSE across a variety of weather situations; for SD and DD models, the percentage improvements vary from 8.07% to 90.29%.

A parametric approach for creating finite element models from point clouds for steel superstructure components in steel girder bridges

Finite element (FE) analysis is widely used to evaluate the conditions of existing bridges. However, the creation of high-fidelity FE models (FEM) for existing bridges is often a challenging task. In view of the ability of 3D laser scanning to capture detailed geometrical information, this study proposes a direct scan-to-FEM strategy specifically for steel bridge superstructure and hence to facilitate the creation of high-fidelity FE models for existing steel girder bridges. Parametric representations are proposed to address partial occlusions in the point clouds as well as to reinforce the connectivity and alignment between the superstructure elements. Mesh generation algorithms are developed to automatically convert the parametric representations into all-hexahedron FE mesh assembly that can be used directly for FE analysis without any mesh verification or editing. Experimental validations are carried out to evaluate the effectiveness of proposed method in term of both mesh quality and analysis accuracy.

Robust resiliency-oriented planning of electricity-hydrogen island energy systems under contingency uncertainty

The capability of electricity-hydrogen island energy systems (EH-IESs) to accommodate contingency uncertainties due to common device failures and power interruptions during extreme events remains poorly developed. The present work addresses this issue by proposing a robust resiliency-oriented planning model for EH-IESs comprising alkaline electrolyzer units, proton exchange membrane fuel cells, and gas turbine units under contingency uncertainties. First, a detailed device model of the EH-IES is established with rated, derated, and shutdown states. Second, the uncertainty of device faults in the EH-IES is addressed in a two-stage adaptive robust planning framework, where capacity planning and system operation are optimized in the first and second stages, respectively. Finally, the proposed robust planning model is divided into a master problem and a sub-problem and is then solved using a nested column-and-constraint generation algorithm. The improved economy and resilience provided by the proposed approach are demonstrated via the results of numerical computations.

Robotic navigation with deep reinforcement learning in transthoracic echocardiography

Abstract Purpose The search for heart components in robotic transthoracic echocardiography is a time-consuming process. This paper proposes an optimized robotic navigation system for heart components using deep reinforcement learning to achieve an efficient and effective search technique for heart components. Method The proposed method introduces (i) an optimized search behavior generation algorithm that avoids multiple local solutions and searches for the optimal solution and (ii) an optimized path generation algorithm that minimizes the search path, thereby realizing short search times. Results The mitral valve search with the proposed method reaches the optimal solution with a probability of 74.4%, the mitral valve confidence loss rate when the local solution stops is 16.3% on average, and the inspection time with the generated path is 48.6 s on average, which is 56.6% of the time cost of the conventional method. Conclusion The results indicate that the proposed method improves the search efficiency, and the optimal location can be searched in many cases with the proposed method, and the loss rate of the confidence in the mitral valve was low even when a local solution rather than the optimal solution was reached. It is suggested that the proposed method enables accurate and quick robotic navigation to find heart components.

Automatic Test Pattern Generation for Robust Quantum Circuit Testing

Quantum circuit testing is essential for detecting potential faults in realistic quantum devices, while the testing process itself also suffers from the inexactness and unreliability of quantum operations. This article alleviates the issue by proposing a novel framework of automatic test pattern generation (ATPG) for robust testing of logical quantum circuits. We introduce the stabilizer projector decomposition (SPD) for representing the quantum test pattern and construct the test application (i.e., state preparation and measurement) using Clifford-only circuits, which are rather robust and efficient as evidenced in the fault-tolerant quantum computation. However, it is generally hard to generate SPDs due to the exponentially growing number of the stabilizer projectors. To circumvent this difficulty, we develop an SPD generation algorithm, as well as several acceleration techniques that can exploit both locality and sparsity in generating SPDs. The effectiveness of our algorithms are validated by (1) theoretical guarantees under reasonable conditions and (2) experimental results on commonly used benchmark circuits, such as Quantum Fourier Transform (QFT), Quantum Volume (QV), and Bernstein-Vazirani (BV) in IBM Qiskit.

An efficient physics-dimension-based reduction method for computing frequency response functions of viscoelastically damped systems

The frequency response functions (FRFs) are of critical interest to dynamic problems, however, they suffer from computational challenges for viscoelastically damped systems. In this paper, a physics-space-based reduction method is proposed for predicting the FRFs of large-scale viscoelastically damped systems involving the standard linear solid model. A physics-dimension subspace is constructed based on original system matrices and viscoelastic parameters, which can be easily generated by using a recursive manner. A projection basis generation algorithm is then developed to generate a standard orthonormal basis within the physics-dimension subspace. With the help of the standard orthonormal basis and the moment-matching-based reduction method, a physics-space-based reduction method is proposed for efficiently predicting the FRFs of large-scale viscoelastically damped systems. Unlike the widely used state-space reduction method, the reduced system of the proposed method can preserve system’s physical structure so that the physical meaning can be captured. Using both theoretical and numerical analyses, the proposed physics-space-based method is more accurate and efficient than the state-space-based reduction method.

A dosimetrically motivated pathfinding approach for non-isocentric dynamic trajectory radiotherapy

Objective.Non-isocentric dynamic trajectory radiotherapy (DTRT) involves dynamic table translations in synchrony with intensity modulation and dynamic gantry, table, and/or collimator rotation. This work aims to develop and evaluate a novel dosimetrically motivated path determination technique for non-isocentric DTRT.Approach.The path determination considers all available beam directions, given on a user-specified grid of gantry angle, table angle, and longitudinal, vertical, and lateral table position. Additionally, the source-to-target distance of all beam directions can be extended by moving the table away from the gantry along the central beam axis to increase the collision-free space. The path determination uses a column generation algorithm to iteratively add beam directions to paths until a user-defined total path length is reached. A subsequent direct aperture optimization of the intensity modulation along the paths creates deliverable plans. Non-isocentric DTRT plans using the path determination and using a manual path setup were created for a craniospinal and a spinal irradiation case. Furthermore, VMAT, isocentric DTRT, and non-isocentric DTRT plans are created for a breast, head and neck (H&N), and esophagus case. Additionally, a HyperArc plan is created for the H&N case. The plans are compared in terms of the dosimetric treatment plan quality and estimated delivery time.Main results.For the craniospinal and spinal irradiation case, using path determination results in dose distributions with improved conformity but a slightly worse target homogeneity compared to manual path setup. The non-isocentric DTRT plans maintained target coverage while reducing the mean dose to organs-at-risk on average by 1.7 Gy (breast), 1.0 Gy (H&N), and 1.6 Gy (esophagus) compared to the VMAT plans and by 0.8 Gy (breast), 0.6 Gy (H&N), and 0.8 Gy (esophagus) compared to the isocentric DTRT plans.Significance.A general dosimetrically motivated path determination applicable to non-isocentric DTRT plans is successfully developed, further advancing the treatment planning for non-isocentric DTRT.

A Comparative Study of Maze Generation Algorithms in a Game-Based Mobile Learning Application for Learning Basic Programming Concepts

This study evaluates several maze generation algorithms applied to generate mazes in a game-based Android mobile application designed to support children in learning basic programming concepts and computational thinking. Each algorithm is assessed for its ability to generate solvable and educationally effective mazes, varying in complexity and size. Key findings indicate that Wilson’s and Aldous–Broder algorithms were identified as the most time inefficient. In comparison, Sidewinder and Binary Tree algorithms perform best for smaller mazes due to their straightforward traversal methods. The Hunt-and-Kill and Recursive backtracker algorithms maintain higher ratios of longest paths, making them suitable for the more complex maze generation required for advanced game levels. Additionally, the study explores various maze-solving algorithms, highlighting the efficiency of the recursive algorithm for simpler mazes and the reliability of Dijkstra’s algorithm across diverse maze structures. This research underscores the importance of selecting appropriate maze generation and solving algorithms to balance generation speed, path complexity, and navigational characteristics. While the study demonstrates the practical applicability of these algorithms in a mobile educational application, it also identifies limitations and suggests directions for future research.

Generating 3D Models for UAV-Based Detection of Riparian PET Plastic Bottle Waste: Integrating Local Social Media and InstantMesh

In recent years, waste pollution has become a severe threat to riparian environments worldwide. Along with the advancement of deep learning (DL) algorithms (i.e., object detection models), related techniques have become useful for practical applications. This work attempts to develop a data generation approach to generate datasets for small target recognition, especially for recognition in remote sensing images. A relevant point is that similarity between data used for model training and data used for testing is crucially important for object detection model performance. Therefore, obtaining training data with high similarity to the monitored objects is a key objective of this study. Currently, Artificial Intelligence Generated Content (AIGC), such as single target objects generated by Luma AI, is a promising data source for DL-based object detection models. However, most of the training data supporting the generated results are not from Japan. Consequently, the generated data are less similar to monitored objects in Japan, having, for example, different label colors, shapes, and designs. For this study, the authors developed a data generation approach by combining social media (Clean-Up Okayama) and single-image-based 3D model generation algorithms (e.g., InstantMesh) to provide a reliable reference for future generations of localized data. The trained YOLOv8 model in this research, obtained from the S2PS (Similar to Practical Situation) AIGC dataset, produced encouraging results (high F1 scores, approximately 0.9) in scenario-controlled UAV-based riparian PET bottle waste identification tasks. The results of this study show the potential of AIGC to supplement or replace real-world data collection and reduce the on-site work load.

Summary of DNS traffic filtering trends as a component of modern information systems security

The study analyzes sources related to methods and technologies for DNS (Domain Name System) traffic filtering. Five main directions are identified that are actively used to enhance security at the DNS level. All examined technologies offer improvements in the quality of DNS filtering. It is emphasized that combining different approaches simultaneously can enhance overall security. The summary of research results on DNS traffic security issues indicates certain problems in the quality of the threat intelligence channels used. Therefore, the implementation of AI and LM technologies should enhance the "depth" of extracting useful information about current threats. It is emphasized that the consideration of information security issues should be conducted exclusively in the context of preventing the disparity of artificial intelligence (AI) capabilities in favor of the adversary (i.e., cybercriminals). Practically, this means that future DNS filtering systems should widely implement the latest advancements in VR, AI, LM, and DL technologies. This is particularly important in countering Domain Generation Algorithm (DGA) mechanisms and the spread of botnets. The specific issues of ensuring a consensus on the security and performance of current information and communication systems when implementing DNS encryption tools are highlighted. The primary problem associated with DNS traffic encryption is the potential for its misuse by attackers to conceal their destructive activities (phishing, spam, etc.).

A finite element analysis method for random fiber-aggregate 3D mesoscale concrete based on the crack bridging law

Mesoscale modeling is an emerging analytical method that can be used to study fiber-reinforced concrete. However, fine fibers with micrometer diameters are distributed in millions in a standard specimen, so it is impossible to establish a mesoscale model and further analyze it, which severely limits the investigation of the reinforcement mechanism of fine fibers on concrete. This paper efficiently established a highly applicable mesoscale analytical model for random fiber-polyhedral aggregate concrete in Abaqus software based on a comprehensive examination of the crack bridging law and the supplementation of the fiber spacing theory, combined with the improvement of the algorithms of aggregate generation, fiber arrangement, and the judgment of the fiber-aggregate contact by the secondary development method of Python, as well as the matrix-dual arrangement method. Additionally, basalt fiber concrete and mortar specimens were prepared and tested, and the parameters of the model were calibrated to verify the model's authenticity. The result shows that the method proposed in this paper can accurately and efficiently perform mesoscale concrete analysis with fine fibers. In the peak load stage, the fiber has the strongest alleviation effect (Δ = 0.033) on the matrix damage, 10 times that of the crack initiation stage (Δ = 0.003). With the development of damage, fiber stress increases gradually from the crack initiation stage (0.68 σfu), and the growth rate (41.53 %) is the largest at the peak load stage (σfu), which reveals the complex dynamic bridging effect of fiber on the matrix. This investigation proposes a novel and broad applicability methodology for the finite element analysis of fiber-reinforced aggregate mesoscale concrete.

Human–object interaction detection algorithm based on graph structure and improved cascade pyramid network

Aiming at the problem of insufficient use of human–object interaction (HOI) information and spatial location information in images, we propose a human–object​ interaction detection network based on graph structure and improved cascade pyramid. This network is composed of three branches, namely, graph branch, human–object branch and human pose branch. In graph branch, we propose a Graph-based Interactive Feature Generation Algorithm (GIFGA) to address the inadequate utilization of interaction information. GIFGA constructs an initial dense graph model by taking humans and objects as nodes and their interaction relationships as edges. Then, by traversing each node, the graph model is updated to generate the final interaction features. In human pose branch, we propose an Improved Cascade Pyramid Network (ICPN) to tackle the underutilization of spatial location information. ICPN extracts human pose features and maps both the object bounding boxes and extracted human pose maps onto the global feature map to capture the most discriminative interaction-related region features within the global context. Finally, the features from the three branches are fed into a Multi-Layer Perceptron (MLP) for fusion and then classified for recognition. Experimental results demonstrate that our network achieves mAP of 54.93% and 28.69% on the V-COCO and HICO-DET datasets, respectively.