Distributed Artificial Intelligence

A Modern Approach

Satya Prakash Yadav, Dharmendra Prasad Mahato, Nguyen Thi Dieu Linh, Satya Prakash Yadav, Dharmendra Prasad Mahato, Nguyen Thi Dieu Linh

• Publisher: CRC Press
• Year: 2020
• ISBN:
  9781000262117
• Edition: 1
• Topic: Ciencia de la computación
• Subtopic: Inteligencia artificial (IA) y semántica

Distributed Artificial Intelligence (DAI) came to existence as an approach for solving complex learning, planning, and decision-making problems. When we talk about decision making, there may be some meta-heuristic methods where the problem solving may resemble like operation research. But exactly, it is not related completely to management research. The text examines representing and using organizational knowledge in DAI systems, dynamics of computational ecosystems, and communication-free interactions among rational agents. This publication takes a look at conflict-resolution strategies for nonhierarchical distributed agents, constraint-directed negotiation of resource allocations, and plans for multiple agents.

Topics included plan verification, generation, and execution, negotiation operators, representation, network management problem, and conflict-resolution paradigms. The manuscript elaborates on negotiating task decomposition and allocation using partial global planning and mechanisms for assessing nonlocal impact of local decisions in distributed planning.

The book will attract researchers and practitioners who are working in management and computer science, and industry persons in need of a beginner to advanced understanding of the basic and advanced concepts.

1 Distributed Artificial Intelligence

Annu Mishra

Contents

1.1 Introduction

1.2 Why Distributed Artificial Intelligence?

1.3 Characteristics of Distributed Artificial Intelligence

1.4 Planning of DAI Multi-Agents

1.5 Coordination among Multi-Agents

1.5.1 Forestalling Mobocracy or Confusion

1.5.2 Meeting Overall Requirements

1.5.3 Distributed Skill, Resources, and Data

1.5.4 Dependency among the Agents

1.5.5 Efficiency

1.6 Communication Modes among the Agents

1.7 Categories of RPC

1.8 Participation of Multi-Agents

1.8.1 Fully Cooperative Architecture

1.8.2 Partial Cooperative Architecture

1.9 Applications of DAI

1.9.1 Electricity Distribution

1.9.2 Telecommunications Systems

1.9.3 Database Technologies for Service Order Processing

1.9.3.1 Concurrent Engineering

1.9.3.2 Weather Monitoring

1.9.3.3 Intelligent Traffic Control

1.10 Conclusion

References

1.1 Introduction

Evolutionary computing has been extensively investigated under varying environmental conditions, particularly with respect to interaction and controlling multiple agents. These agents can be autonomous bodies, such as software programs, applications, or robots. Multi-agent architecture may be used to investigate phenomena or to remedy issues that are difficult for human beings to examine and clear up. They may be utilized in various areas, ranging from computer games and informatics to economics and social sciences (Varro et al. 2005; Sun and Naveh 2004; Gutnik and Kaminka 2004; Kubera et al. 2010). Evolutionary computing has evolved as a paradigm for complex and composite problems. The basic idea is involvement of multi-agent to solve a problem instead of a single agent problem-solving technique. These agents are capable of discovering the solution by making their own decisions in the presence of multiple agents (Shoham and Leyton-Brown 2010). Traditional Artificial Intelligence (AI) systems have concentrated on gathering information, knowledge representation, and execution via a single intelligent agent, whereas in distributed artificial intelligence (DAI), these processes are distributed among the agents that are proficient for making independent as well as coordinated decisions.

According to Ponomarev and Voronkov (2017), a DAI can be characterized by three principle qualities (Sichman et al. 1994):

1. 1. It is a technique for the dispersion of jobs between operators;
2. 2. It is a technique for dispersion of forces;
3. 3. It is a technique for communicating among the participating agents.

To satisfy the first quality, two aspects must be considered: first, the dissemination of knowledge and, second, the allocation of tasks among the agents.

1.2 Why Distributed Artificial Intelligence?

Modeling and computational responsibilities have gained extra complexity as dimensions continue to grow. As a result, it is hard to address the use of centralized methods. Although motivations to apply multi-agent systems (MASs) for researchers from various disciplines are special, as discussed by Yu and Liu (2016), the principal benefits of using multi-agent technologies include (1) individuals or agents keep track of software-specific nature and environment; (2) interactions among agents can be modeled and supervised; and (3) any difficulty is managed by sublayers and/or subcomponents of the system. Therefore, MASs can be considered to be an optimal solution for a computational paradigm in a distributed environment as well as for vast data complexities. In addition, AI techniques can be utilized.

Pattison et al. (1987) state that agents take sensory input from the environment produces the output action as shown in Figure 1.1.

FIGURE 1.1 Distributed artificial intelligence structure. Source: Pattison et al. (1987).

Before we study multi-agents further, let us first understand the term “Agent.” An agent has the following characteristics:

• Self-sufficiency: agents work without the immediate mediation of people or others, and have an authority over their activities and inside state.
• Social capacity: agents collaborate with different agents (and conceivably people) through a specialist correspondence language.
• Reactivity: agents see their condition (which might be the physical world, a client by means of a graphical UI, an assortment of different specialists, the Internet, or maybe these joined), and react in an auspicious design to changes that happen in it.
• Pro-animation: agents do not just act in light of their condition; they can show objective coordinated conduct by stepping up to the plate.

The independent operator approach replaces a concentrated database and control PC with a system of operators, each supplied with a neighborhood perspective on its condition and the capacity and power to react locally to that condition (Andrews 1991). The general framework execution is not all inclusive arranged; however, it develops progressively through the dynamic cooperation of specialists. Hence, the framework does not shift back and forth between patterns of booking and execution. Thus, the computation rises up out of the simultaneous free choices of the neighborhood operators or the multi-agents.

A distributed approach additionally offers benefits when managing vulnerability, for example in the area of information. In open frameworks (Hewitt and de Jong 1983), it is frequently beneficial to have numerous central focuses of critical thinking action so as to have the option to manage more than one unexpected occasion at once. A distributed framework will not experience the ill effects of single point failure. This gives it a preferred position in circumstances where unwavering quality is of specific significance. In a distributed framework, the failure of a critical thinking hub or a single point will bring about an effortless debasement of execution instead of a total framework disappointment. Likewise, from a framework structure viewpoint, an appropriated arrangement might be the most practical. It might be that the all-out expense of the quantity of minimal effort equipment gadgets required to actualize a disseminated framework is lower than the expense of a huge gadget that is groundbreaking enough to play out a similar assignment. It might likewise be the situation that the measured quality upheld by a disseminated approach lessens programming costs. DAI leads to distributed problem-solving. This approach tackles a specific issue that can be separated among various modules that coordinate in isolating and sharing information about the issue and its solution(s). Distributed problem-solving is a type of critical thinking that manages the issues by disseminating it among several problem solvers which communicate with each other in a cooperative manner and share the burden of transferring the partial solution to each other. The communication between partial problem solvers is expressly characterized in appropriated critical thinking situations. Multi-agent systems, by contrast, works together with loosely coupled problem solvers to tackle an issue. These autonomous problem solvers, or computational operators, are self-sufficient elements that have the capacity to brilliantly work under different natural conditions given their tangible and strong abilities (Lesser 1991). Figure 1.2 depicts the phases or steps of a solver. It can be observed that, first, the problem that arrives is fragmented into several sections. Fragmentation continues until the unit issue is obtained. In the next phase, solvers try to find a solution for each unit issue by interacting and sharing information and amenities. In the last and final phase, the subsolutions obtained are integrated to form a final solution to a problem.

FIGURE 1.2 Phases of solving a problem in distributed environment.

1.3 Characteristics of Distributed Artificial Intelligence

1. 1. Distributed artificial intelligence (DAI), also called decentralized artificial intelligence, is a subfield of artificial intelligence (AI) that is committed to the improvement of distributed reasoning for issues.
2. 2. DAI is firmly identified by multi-agent systems and distributed problem solving (DPS).
3. 3. The goals of DAI are to illuminate the thinking, arranging, learning, and recognition issues of AI, particularly when they require enormous amounts of information, by passing off the issue to independent agents.
4. 4. DAI takes into consideration the interconnection and interoperation of different existing hereditaments system. By building an agent sheath around such legacy systems, they can be incorporated into the agents list.
5. 5. DAI upgrades framework system execution, explicitly usi...

Table of contents

• Cover
• Half-Title
• Series
• Title
• Copyright
• Contents
• Preface
• Editors
• Contributors
• Chapter 1 Distributed Artificial Intelligence
• Chapter 2 Intelligent Agents
• Chapter 3 Knowledge-Based Problem-Solving: How AI and Big Data Are Transforming Health Care
• Chapter 4 Distributed Artificial Intelligence for Document Retrieval
• Chapter 5 Distributed Consensus
• Chapter 6 DAI for Information Retrieval
• Chapter 7 Decision Procedures
• Chapter 8 Cooperation through Communication in a Distributed Problem-Solving Network
• Chapter 9 Instantiating Descriptions of Organizational Structures
• Chapter 10 Agora Architecture
• Chapter 11 Test Beds for Distributed AI Research
• Chapter 12 Real-Time Framework Competitive Distributed Dilemma
• Chapter 13 Comparative Studied Based on Attack Resilient and Efficient Protocol with Intrusion Detection System Based on Deep Neural Network for Vehicular System Security
• Chapter 14 A Secure Electronic Voting System Using Decentralized Computing
• Chapter 15 DAI for Document Retrieval
• Chapter 16 A Distributed Artificial Intelligence: The Future of AI
• Chapter 17 Analysis of Hybrid Deep Neural Networks with Mobile Agents for Traffic Management in Vehicular Adhoc Networks
• Chapter 18 Data Science and Distributed AI
• Index