[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
5.3
2.5
Journals
Applied Sciences
Special Issues
Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and...
applsci-logo
Submit to Special Issue Submit Abstract to Special Issue Review for Applied Sciences Propose a Special Issue

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends

A special issue of Applied Sciences (ISSN 2076-3417). This special issue belongs to the section "Computing and Artificial Intelligence".

Deadline for manuscript submissions: 20 May 2025 | Viewed by 10142

Special Issue Editor

Dr. André Sales Mendes

E-Mail Website
Guest Editor
Expert Systems and Applications Lab (ESALAB), Faculty of Science, University of Salamanca, 37008 Salamanca, Spain
Interests: computer languages and systems; computer science
Special Issues, Collections and Topics in MDPI journals

Special Issue Information

Dear Colleagues,

We are pleased to invite you to contribute to our upcoming Special Issue on multi-agent systems, a crucial and growing area of research in computer science and artificial intelligence (AI). These systems consist of multiple interacting agents that can work collaboratively or competitively to achieve individual or collective goals. The study of multi-agent systems is vital due to their applications in various fields, including robotics, automated negotiation, distributed problem solving, among others.

This Special Issue aims to advance the understanding and development of multi-agent systems by exploring novel methodologies, applications, and theoretical foundations, particularly focusing on their integration with AI. The convergence of multi-agent systems and AI has the potential to revolutionize various sectors by enabling intelligent, autonomous, and cooperative solutions to complex problems.

In this Special Issue, original research articles and reviews are welcome. The research areas may include (but are not limited to) the following:

  • Coordination and collaboration mechanisms in multi-agent systems.
  • Applications of multi-agent systems in robotics and automation.
  • Agent-based modeling and simulation techniques.
  • Distributed artificial intelligence.
  • Multi-agent learning and adaptation.
  • Ethical and social implications of multi-agent systems.
  • AI-driven decision making in multi-agent environments.
  • Real-world applications and case studies integrating AI and multi-agent systems.

We encourage submissions that demonstrate novel theoretical insights, practical applications, and interdisciplinary approaches to solving complex problems using multi-agent systems and AI. This Special Issue will serve as a comprehensive resource for researchers, practitioners, and policymakers interested in the future of multi-agent systems and their synergy with artificial intelligence.

Dr. André Sales Mendes
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Applied Sciences is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • multi-agent systems
  • artificial intelligence
  • coordination mechanisms
  • agent-based modeling
  • distributed AI
  • robotics
  • autonomous systems
  • ethical implications
  • AI-driven decision making

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue policies can be found here.

Published Papers (7 papers)

Download All Papers
Order results
Result details
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:

Research

Jump to: Review, Other

22 pages, 3751 KiB  
Article
Bio-Inspired Traffic Pattern Generation for Multi-AMR Systems
by Rok Vrabič, Andreja Malus, Jure Dvoršak, Gregor Klančar and Tena Žužek
Appl. Sci. 2025, 15(5), 2849; https://doi.org/10.3390/app15052849 - 6 Mar 2025
Viewed by 154
Abstract
In intralogistics, autonomous mobile robots (AMRs) operate without predefined paths, leading to complex traffic patterns and potential conflicts that impact system efficiency. This paper proposes a bio-inspired optimization method for autonomously generating spatial movement constraints for autonomous mobile robots (AMRs). Unlike traditional multi-agent [...] Read more.
In intralogistics, autonomous mobile robots (AMRs) operate without predefined paths, leading to complex traffic patterns and potential conflicts that impact system efficiency. This paper proposes a bio-inspired optimization method for autonomously generating spatial movement constraints for autonomous mobile robots (AMRs). Unlike traditional multi-agent pathfinding (MAPF) approaches, which focus on temporal coordination, our approach proactively reduces conflicts by adapting a weighted directed grid graph to improve traffic flow. This is achieved through four mechanisms inspired by ant colony systems: (1) a movement reward that decreases the weight of traversed edges, similar to pheromone deposition, (2) a delay penalty that increases edge weights along delayed paths, (3) a collision penalty that increases weights at conflict locations, and (4) an evaporation mechanism that prevents premature convergence to suboptimal solutions. Compared to the existing approaches, the proposed approach addresses the entire intralogistic problem, including plant layout, task distribution, release and dispatching algorithms, and fleet size. Its autonomous movement rule generation and low computational complexity make it well suited for dynamic intralogistic environments. Validated through physics-based simulations in Gazebo across three scenarios, a standard MAPF benchmark, and two industrial environments, the movement constraints generated using the proposed method improved the system throughput by up to 10% compared to unconstrained navigation and up to 4% compared to expert-designed solutions while reducing the need for conflict-resolution interventions. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>Illustration of the bio-inspired mechanisms. (<b>a</b>) AMR movements and the corresponding weights after applying (<b>b</b>) movement rewards (pheromone deposition), (<b>c</b>) collision handling, (<b>d</b>) delay feedback, and (<b>e</b>) pheromone evaporation.</p> Full article ">Figure 2
<p>Solutions to the test problem: (<b>a</b>) first solution and (<b>b</b>) second solution.</p> Full article ">Figure 3
<p>The emergence of the resulting movement patterns.</p> Full article ">Figure 4
<p>Additional test scenarios demonstrating the impact of different parameters on the emerging movement patterns: (<b>a</b>) solution with a single AMR, (<b>b</b>) solution without collision penalty, and (<b>c</b>) solution with additional tasks.</p> Full article ">Figure 5
<p>Sensitivity analysis of key parameters by phase.</p> Full article ">Figure 6
<p>Comparison of algorithm performance when Phase 1 is removed, means and standard deviations.</p> Full article ">Figure 7
<p>The solution to the rooms’ layout, obtained through the presented algorithm.</p> Full article ">Figure 8
<p>The (<b>a</b>) expert and (<b>b</b>) algorithmic solutions for industrial scenario A. Pickups in light, intermediate buffers in medium, and dropoffs in dark blue.</p> Full article ">Figure 9
<p>The (<b>a</b>) expert and (<b>b</b>) algorithmic solutions for industrial scenario B.</p> Full article ">Figure 10
<p>Analysis of punctuality for industrial scenarios: (<b>a</b>) scenario A and (<b>b</b>) scenario B.</p> Full article ">Figure 11
<p>Simulation in ROS2/Gazebo: (<b>a</b>) 3D view of the rooms’ layout, (<b>b</b>) top-down view, and (<b>c</b>) RViZ view.</p> Full article ">Figure 12
<p>Performance comparison across scenarios. Bars show tasks completed with and without recovery actions; whiskers indicate the standard deviation of total task completions.</p> Full article ">
26 pages, 1579 KiB  
Article
Understanding Recruiters’ Acceptance of Artificial Intelligence: Insights from the Technology Acceptance Model
by Filomena Almeida, Ana Junça Silva, Sara L. Lopes and Isabel Braz
Appl. Sci. 2025, 15(2), 746; https://doi.org/10.3390/app15020746 - 14 Jan 2025
Viewed by 1643
Abstract
The integration of new technologies in professional contexts has emerged as a critical determinant of organizational efficiency and competitiveness. In this regard, the application of Artificial Intelligence (AI) in recruitment processes facilitates faster and more accurate decision-making by processing large volumes of data, [...] Read more.
The integration of new technologies in professional contexts has emerged as a critical determinant of organizational efficiency and competitiveness. In this regard, the application of Artificial Intelligence (AI) in recruitment processes facilitates faster and more accurate decision-making by processing large volumes of data, minimizing human bias, and offering personalized recommendations to enhance talent development and candidate selection. The Technology Acceptance Model (TAM) provides a valuable framework for understanding recruiters’ perceptions of innovative technologies, such as AI tools and GenAI. Drawing on the TAM, a model was developed to explain the intention to use AI tools, proposing that perceived ease of use and perceived usefulness influence attitudes toward AI, which subsequently affect the intention to use AI tools in recruitment and selection processes. Two studies were conducted in Portugal to address this research objective. The first was a qualitative exploratory study involving 100 interviews with recruiters who regularly utilize AI tools in their professional activities. The second study employed a quantitative confirmatory approach, utilizing an online questionnaire completed by 355 recruiters. The qualitative findings underscored the transformative role of AI in recruitment, emphasizing its potential to enhance efficiency and optimize resource management. However, recruiters also highlighted concerns regarding the potential loss of personal interaction and the need to adapt roles within this domain. The results also supported the indirect effect of perceived ease of use and perceived usefulness on the use of AI tools in recruitment and selection processes via positive attitudes toward the use of these tools. This suggests that AI is best positioned as a complementary tool rather than a replacement for human decision-making. The insights gathered from recruiters’ perspectives provide actionable recommendations for organizations seeking to leverage AI in recruitment processes. Specifically, the findings show the importance of ethical considerations and maintaining human involvement to ensure a balanced and effective integration of AI tools. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>Model of intention to use AI tools in Recruitment and Selection.</p> Full article ">Figure 2
<p>Overview of studies.</p> Full article ">Figure 3
<p>Main advantages associated with AI tools in recruitment and selection.</p> Full article ">Figure 4
<p>AI tools disadvantages in recruitment and selection.</p> Full article ">Figure 5
<p>Path coefficients of the proposed mediation model. <span class="html-italic">Note.</span> PU = Perceived usefulness; PEU = Perceived ease of use; AT = Attitude; IU = Intention to use. All path coefficients were significant.</p> Full article ">
40 pages, 1079 KiB  
Article
Context-Adaptable Deployment of FastSLAM 2.0 on Graphic Processing Unit with Unknown Data Association
by Jessica Giovagnola, Manuel Pegalajar Cuéllar and Diego Pedro Morales Santos
Appl. Sci. 2024, 14(23), 11466; https://doi.org/10.3390/app142311466 - 9 Dec 2024
Viewed by 1073
Abstract
Simultaneous Localization and Mapping (SLAM) algorithms are crucial for enabling agents to estimate their position in unknown environments. In autonomous navigation systems, these algorithms need to operate in real-time on devices with limited resources, emphasizing the importance of reducing complexity and ensuring efficient [...] Read more.
Simultaneous Localization and Mapping (SLAM) algorithms are crucial for enabling agents to estimate their position in unknown environments. In autonomous navigation systems, these algorithms need to operate in real-time on devices with limited resources, emphasizing the importance of reducing complexity and ensuring efficient performance. While SLAM solutions aim at ensuring accurate and timely localization and mapping, one of their main limitations is their computational complexity. In this scenario, particle filter-based approaches such as FastSLAM 2.0 can significantly benefit from parallel programming due to their modular construction. The parallelization process involves identifying the parameters affecting the computational complexity in order to distribute the computation among single multiprocessors as efficiently as possible. However, the computational complexity of methodologies such as FastSLAM 2.0 can depend on multiple parameters whose values may, in turn, depend on each specific use case scenario ( ingi.e., the context), leading to multiple possible parallelization designs. Furthermore, the features of the hardware architecture in use can significantly influence the performance in terms of latency. Therefore, the selection of the optimal parallelization modality still needs to be empirically determined. This may involve redesigning the parallel algorithm depending on the context and the hardware architecture. In this paper, we propose a CUDA-based adaptable design for FastSLAM 2.0 on GPU, in combination with an evaluation methodology that enables the assessment of the optimal parallelization modality based on the context and the hardware architecture without the need for the creation of separate designs. The proposed implementation includes the parallelization of all the functional blocks of the FastSLAM 2.0 pipeline. Additionally, we contribute a parallelized design of the data association step through the Joint Compatibility Branch and Bound (JCBB) method. Multiple resampling algorithms are also included to accommodate the needs of a wide variety of navigation scenarios. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>FastSLAM 2.0 pipeline.</p> Full article ">Figure 2
<p>Observation model—graphical representation.</p> Full article ">Figure 3
<p>Hardware–software architecture schema.</p> Full article ">Figure 4
<p>Simulation environment schema.</p> Full article ">Figure 5
<p>Functional blocks partitioning schema.</p> Full article ">Figure 6
<p>Detailed heterogeneous architecture pipeline.</p> Full article ">Figure 7
<p>Data association pipeline.</p> Full article ">Figure 8
<p>Particle Initialization—elapsed time.</p> Full article ">Figure 9
<p>Particle Prediction—elapsed time.</p> Full article ">Figure 10
<p>Mahalanobis Distance—elapsed time.</p> Full article ">Figure 11
<p>Problem Preparation—elapsed time.</p> Full article ">Figure 12
<p>Branch and Bound—elapsed time.</p> Full article ">Figure 13
<p>Proposal Adjustment—elapsed time.</p> Full article ">Figure 14
<p>Landmark Estimation—elapsed time.</p> Full article ">Figure 15
<p>Resampling traditional methods—elapsed time.</p> Full article ">Figure 16
<p>Resampling alternative methods—elapsed time.</p> Full article ">
15 pages, 474 KiB  
Article
Federated Learning in Dynamic and Heterogeneous Environments: Advantages, Performances, and Privacy Problems
by Fabio Liberti, Davide Berardi and Barbara Martini
Appl. Sci. 2024, 14(18), 8490; https://doi.org/10.3390/app14188490 - 20 Sep 2024
Viewed by 2821
Abstract
Federated Learning (FL) represents a promising distributed learning methodology particularly suitable for dynamic and heterogeneous environments characterized by the presence of Internet of Things (IoT) devices and Edge Computing infrastructures. In this context, FL allows you to train machine learning models directly on [...] Read more.
Federated Learning (FL) represents a promising distributed learning methodology particularly suitable for dynamic and heterogeneous environments characterized by the presence of Internet of Things (IoT) devices and Edge Computing infrastructures. In this context, FL allows you to train machine learning models directly on edge devices, mitigating data privacy concerns and reducing latency due to transmitting data to central servers. However, the heterogeneity of computational resources, the variability of network connections, and the mobility of IoT devices pose significant challenges to the efficient implementation of FL. This work explores advanced techniques for dynamic model adaptation and heterogeneous data management in edge computing scenarios, proposing innovative solutions to improve the robustness and efficiency of federated learning. We present an innovative solution based on Kubernetes which enables the fast application of FL models to Heterogeneous Architectures. Experimental results demonstrate that our proposals can improve the performance of FL in IoT and edge environments, offering new perspectives for the practical implementation of decentralized intelligent systems. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>Schema of implementation using k3s. The k3s worker can be any architecture compatible with Kubernetes, with heterogeneous architecture. For instance, in our implementation, k3s workers are ARM virtual machines running in a cloud provider (Oracle).</p> Full article ">Figure 2
<p>Analysis of the performance of the system over 150 epochs in terms of accuracy and loss over five different datasets and with clients varying from two to fifty.</p> Full article ">Figure 3
<p>Comparison of the accuracy of federated and centralized learning systems with similar works.</p> Full article ">
22 pages, 5394 KiB  
Article
Target-Oriented Multi-Agent Coordination with Hierarchical Reinforcement Learning
by Yuekang Yu, Zhongyi Zhai, Weikun Li and Jianyu Ma
Appl. Sci. 2024, 14(16), 7084; https://doi.org/10.3390/app14167084 - 12 Aug 2024
Viewed by 1722
Abstract
In target-oriented multi-agent tasks, agents collaboratively achieve goals defined by specific objects, or targets, in their environment. The key to success is the effective coordination between agents and these targets, especially in dynamic environments where targets may shift. Agents must adeptly adjust to [...] Read more.
In target-oriented multi-agent tasks, agents collaboratively achieve goals defined by specific objects, or targets, in their environment. The key to success is the effective coordination between agents and these targets, especially in dynamic environments where targets may shift. Agents must adeptly adjust to these changes and re-evaluate their target interactions. Inefficient coordination can lead to resource waste, extended task times, and lower overall performance. Addressing this challenge, we introduce the regulatory hierarchical multi-agent coordination (RHMC), a hierarchical reinforcement learning approach. RHMC divides the coordination task into two levels: a high-level policy, assigning targets based on environmental state, and a low-level policy, executing basic actions guided by individual target assignments and observations. Stabilizing RHMC’s high-level policy is crucial for effective learning. This stability is achieved by reward regularization, reducing reliance on the dynamic low-level policy. Such regularization ensures the high-level policy remains focused on broad coordination, not overly dependent on specific agent actions. By minimizing low-level policy dependence, RHMC adapts more seamlessly to environmental changes, boosting learning efficiency. Testing demonstrates RHMC’s superiority over existing methods in global reward and learning efficiency, highlighting its effectiveness in multi-agent coordination. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>Regulatory hierarchical multi-agent coordination (RHMC) architecture.</p> Full article ">Figure 2
<p>The frame of high-level.</p> Full article ">Figure 3
<p>The frame of low-level.</p> Full article ">Figure 4
<p>The model of regulation algorithm.</p> Full article ">Figure 5
<p>Target-oriented multi-agent cooperative environments: cooperative navigation.</p> Full article ">Figure 6
<p>The learning curve of our method with baselines in cooperative navigation.</p> Full article ">Figure 7
<p>Learning Curves of RHMC and Baselines in Multi-Sensor Multi-Target Coverage Task.</p> Full article ">Figure 8
<p>Experimental environment with different number of sensors with five targets in multi-sensor multi-target coverage Task.</p> Full article ">Figure 9
<p>The result of our method with baselines in different number of sensors with five targets.</p> Full article ">Figure 10
<p>The learning curve of our method with its ablations in multi-sensor multi-target coverage Task.</p> Full article ">

Review

Jump to: Research, Other

22 pages, 1206 KiB  
Review
A Systematic Review of Intelligent Systems and Analytic Applications in Credit Card Fraud Detection
by Ercan Oztemel and Muhammed Isik
Appl. Sci. 2025, 15(3), 1356; https://doi.org/10.3390/app15031356 - 28 Jan 2025
Viewed by 836
Abstract
The use of credit cards plays a crucial role in cash management and in meeting the needs for individual and commercial customers due to the spread of risks to the future by making monthly instalments instead of cash transactions. The use of credit [...] Read more.
The use of credit cards plays a crucial role in cash management and in meeting the needs for individual and commercial customers due to the spread of risks to the future by making monthly instalments instead of cash transactions. The use of credit cards therefore provides benefits not only to the customers but also to the banks as it enables and sustains a long-term relationship in between them. Despite the increase in the use of credit cards, there is also a significant increase in fraud transactions. To detect and prevent possible fraud operations, banks generally use rule-based techniques or analytical models. In this respect, analytical models have an important place due to their effectiveness, performance, and fast response. The main aim of this paper is therefore to enhance the theoretical and practical understanding of credit card fraud operations, review basic approaches, and propose a more comprehensive approach utilizing the agents. Note that in this study, static analytic modelling (existing approaches) and dynamic analytic modelling (emerging approaches) techniques are compared in terms of methodology, performance, and respective approaches. Since fraud methods and transactions are constantly changing over time, it is thought that there will be an increase in the use of agent-based models with dynamic analytical capabilities. Additionally, in this paper, a proposed model and empiric study are presented for an agent-based intelligent credit card fraud detection system. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>Credit card fraud attacks.</p> Full article ">Figure 2
<p>Proposed conceptual RL model in detecting credit card fraud.</p> Full article ">

Other

Jump to: Research, Review

36 pages, 988 KiB  
Systematic Review
A Systematic Review on Reinforcement Learning for Industrial Combinatorial Optimization Problems
by Miguel S. E. Martins, João M. C. Sousa and Susana Vieira
Appl. Sci. 2025, 15(3), 1211; https://doi.org/10.3390/app15031211 - 24 Jan 2025
Viewed by 742
Abstract
This paper presents a systematic review on reinforcement learning approaches for combinatorial optimization problems based on real-world industrial applications. While this topic is increasing in popularity, explicit implementation details are not always available in the literature. The main objective of this paper is [...] Read more.
This paper presents a systematic review on reinforcement learning approaches for combinatorial optimization problems based on real-world industrial applications. While this topic is increasing in popularity, explicit implementation details are not always available in the literature. The main objective of this paper is characterizing the agent–environment interactions, namely, the state space representation, action space mapping and reward design. Also, the main limitations for practical implementation and the needed future developments are identified. The literature selected covers a wide range of industrial combinatorial optimization problems, found in the IEEE Xplore, Scopus and Web of Science databases. A total of 715 unique papers were extracted from the query. Then, out-of-scope applications, reviews, surveys and papers with insufficient implementation details were removed. This resulted in a total of 298 papers that align with the focus of the review with sufficient implementation details. The state space representation shows the most variety, while the reward design is based on combinations of different modules. The presented studies use a large variety of features and strategies. However, one of the main limitations is that even with state-of-the-art complex models the scalability issues of increasing problem complexity cannot be fully solved. No methods were used to assess risk of biases or automatically synthesize the results. Full article
(This article belongs to the Special Issue Advancements in Multi-Agent Systems and Artificial Intelligence: Methodologies, Applications, and Future Trends)
►▼ Show Figures

Figure 1

Figure 1
<p>Reinforcement learning loop representation, based on [<a href="#B1-applsci-15-01211" class="html-bibr">1</a>].</p> Full article ">Figure 2
<p>PRISMA flow diagram of query results breakdown, made using [<a href="#B15-applsci-15-01211" class="html-bibr">15</a>].</p> Full article ">Figure 3
<p>Distribution by publication date. The earliest paper is from 2002.</p> Full article ">Figure 4
<p>Algorithm type distribution.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-7
Back to TopTop