26
Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Embed Size (px)

Citation preview

Page 1: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Bijan Zamanian and Zeki Bayram

Eastern Mediterranean University Department of Computer Engineering

Page 2: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Introduction What is an agent? What is coordination? Why coordination Coordination models Our model Proposed architecture Demonstration of proposed architecture Conclusion

2

Page 3: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Complex tasks are often carried out by teams consisting of individuals, because no one individual has the collective expertise, information, or resources required for the effective completion or performance of a task.

Agents can cooperate to facilitate achieving a common, complicated and large scale goal using some of their characteristic like Intelligence , autonomy. In such a case

each agent is responsible for a part of the goal .  But to ensure a community of individual agents acts in a coherent manner they need coordination.  Coordination may require cooperation, but cooperation among a set of agents does not necessarily results in coordination.

3

Page 4: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1) A software agent is a piece of software that acts on behalf of a user or other programs.

2) An agent is defined in terms of its behavior.3) Agents commonly include the following

concepts persistence (code is not executed on demand but runs continuously

and decides for itself when it should perform some activity) autonomy (agents have capabilities of task selection, prioritization,

goal-directed behavior, decision-making without human intervention) sociability (agents are able to engage other components through

some sort of communication and coordination, they may collaborate on a task)

reactivity (agents perceive the context in which they operate and react to it appropriately).

Intelligence (agents are capable of reasoning and learning )

4

Page 5: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

4. A cooperative system of agents may fall into one or more of these categories : distributed agents (being executed on physically distinct

computers), multi-agent systems (distributed agents that do not have the

capabilities to achieve an objective alone and thus must communicate),

mobile agents (agents that can relocate their execution onto different processors).

5

Page 6: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1. Coordination in general : Act of making different people or things work together for a goal or

effect.

2. Agent coordination.The process by which agents reason about their local actions and the

(anticipated) actions of others .

6

Page 7: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

7

Page 8: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1. Coordination model: An agent coordination model is a conceptual framework .

2. Coordination architecture: A coordination architecture is a software infrastructure.

3. Some well known coordination models are: Direct, Meeting oriented, Blackboard-based Linda-like

8

Page 9: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1. DirectIn Direct coordination models, agents usually coordinate using RPC-like

primitives or synchronous message passing.

2. Meeting orientedIn Meeting oriented models, agents coordinate using implicit or known

meeting points.

3. Blackboard-basedIn Blackboard-based models, agents coordinate via shared data spaces

to store and retrieve information under the form of messages .

4. Linda-likeIn Linda-like models, agents coordinate through tuple spaces which

allow for Insertion of tuples and retrieval of tuples using associative pattern

matching.

9

Page 10: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1. Dynamics Coordination strategies that handle highly dynamic task environments

may not keep up with all environment

2. Agent population propertiesA coordinator (specially centralized ) can quickly degrade and

becomes incapable of processing the interactions if population of agents increases.

3. Quantity of interactionIf each agent interacts with every other agent, the number of paired

interactionswill grow quadratically

10

Page 11: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1. Our model can be seen as an instance of blackboard-based coordination.

2. Our model obviates some coordination concerns (Dynamics, Agent population properties, Quantity of interaction)

3. Hierarchical control of agents4. Actual reality vs. perceived reality (agents and

coordinator may not have full or perfect view of reality)

5. Web services: eyes, ears of agents. Also effectors of agents. Has full view of reality, but reveals only part.

6. Coordinator: one for each group of agents

11

Page 12: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Web service

12

Page 13: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Our architecture is made of four main parts: Coordinator, Agents, Environment and Web service.

Web Service

EJB Coordinator

send /receive tasksend/ receive task result

perceive /effect/ query environment

Environment

14

Page 14: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Agents implemented as Java objects running in their own thread

Agent

x,y,z

runner

Miss.1

health

ammo

type

mid

Miss.2

Run()

Askmission()

DoMission()

RI_lookup()

Observe()

IsAccomplished() Go()

16

Page 15: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Mission request from superior (coordinator)

Web services used to be informed about environment, and cause changes to environment

No direct inter-agent communication Report result of mission to coordinator Built-in intelligence: observation and

decision making17

Page 16: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

18

Page 17: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Coordinators implemented as EJB’s (Enterprise Java Beans

Responsible for coordinating a group of agents

Agents in the field communicate with them to get initial mission,and to report result of mission

A different set of coordinators for each “team” possible in case of multi-part simulation

19

Page 18: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

1. Logic of mission creationCoordinator generates missions for units its under command based on

the units’ types and enemy types

2. How it perceivesInitially the coordination does not know about the presence of the

enemy units. Initial perception through agent observations

20

Page 19: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Web Service

AgendDb

Data Source

Method 1 Method 2 Method n

Ws-Method 1

Ws-Method 2

Ws-Method n

21

Page 20: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Web services maintain the actual reality Agents call methods to inquire about their

environment (who is around me?) Agents call methods to change

environment (shoot the guy) Web service methods reveal only part of

the reality, given the situation of the requesting agent. Actions are carried out in a probabilistic manner

22

Page 21: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Agents can exist in different java virtual machines or even on different physical machines.

Coordinators can be distributed on different machines Job of coordination made easy through coordination

hierarchy. One simulation can run on multiple machines and

share the same “environment” through using Web services. Different “teams” (e.g. enemies) can run in different machines, but participate in the same simulation

Our architecture is both extendable and scalable

23

Page 22: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Modeled armed forces:

WS- Coordinator

24

Page 23: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Described the agent coordination problem

Defined an agent coordination architecture based on EJB’s and WS’s

Demonstrated the architecture in a battlefield simulation

Our architecture has nice properties: scalable, extendable

25

Page 24: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Web Service

Environment

EJB Coordinator

34

Page 25: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Dieter Fensel;Holger Lausen;Jos de Bruijn;Michael Stollberg;Dumitru Roman (2007):"Enabling Semantic Web Services". The Web Service Modeling Ontology. Springer Berlin Heidelberg Publishing. ISBN 978-3-540-34519-0 (Print) 978-3-540-34520-6 (Online).

H S Nwana, L Lee and N R Jennings, (1996): “Co-ordination in software agent systems.” BT Technol J Vol 14 No 4 October 1996.

Stollberg, M.; Haller, A.(2005): "Services Computing", 2005 IEEE International Conference on Volume 2,  11-15 July 2005 Page(s):xv vol.2

Stollberg, M.; Haller, A. (2005):" Semantic Web services tutorial", Mar-Apr 2001Volume: 2, Issue: 2 On page(s): On page(s): xv vol.2 ,Number of Pages: 2 vol. (xxi+660) ,ISBN: 0-7695-2408-7 INSPEC Accession Number:8652031 ,,Date Published in Issue: 2005-11-21 08:54:29

Weiming Shen; Hamada Ghenniwa; Yinsheng Li (2006): "Agent-Based Service-Oriented Computing and Applications", Pervasive Computing and Applications, 2006 1st International Symposium on.3-5 Aug. 2006 Page(s):8 – 9.

35

Page 26: Bijan Zamanian and Zeki Bayram Eastern Mediterranean University Department of Computer Engineering

Noh-sam Park; Gil-haeng Lee (2003):"Agent-based Web services middleware", Global Telecommunications Conference, 2003. GLOBECOM '03. IEEE Volume 6,  1-5 Dec. 2003 Page(s):3186 - 3190 vol.6

Zhi-Zhong Sun; Bin Li; Liang Li (2007):"An Adaptive Agent Coordination Framework for Web Services Composition", Machine Learning and Cybernetics, 2007 International Conference on Volume 7,  19-22 Aug. 2007 Page(s):3870 - 3875

R. Scott Cost, Yannis K Labrou,Tim Finin:"Agent Communication Languages and Agent Coordination", Coordination of Internet Agents: Models, Technologies and Applications. July 01, 2000.

Willmott, S.; Pena, F.O.F.; Merida-Campos, C.; Constantinescu, I.; Dale, J.; Cabanillas, D. (2005): "Adapting agent communication languages for semantic Web service inter-communication", Web Intelligence, 2005. Proceedings. The 2005 IEEE/WIC/ACM International Conference on

19-22 Sept. 2005 Page(s):405 – 408

36