Software Agents

1
Software Agents

• Gabriel Becerra Knowledge Science Group
  ayala_at_cpsc.ucalgary.cawww.cpsc.ucalgary.ca/ayal
  a
• Computer SciencesUniversity of Calgary

2
Todays Agenda

• Agents
• Multi-Agent Systems
• Standards
• Agent Communication Languages
• Example
• Concluding Remarks
• Questions

3
Agents
4
Agent Definition (1/1)

• Autonomous Process capable to react, initiate
  changes, establish communications, etc. (Jennings
  and Woolridge 1998)
• Agents are composed of
• Set of situations (Sit)
• Set of actions (Act)
• Set of internal data (Data)

FAG Sit x Data ? Act
Mobile really?
5
Multi-Agent Systems
6
Multi-agent Systems (1/6)

• Describe several agents that interact with each
  other
• A multi-agent system is composed of
• A set of situations (sit)
• A set of actions (Mact)
• A set of agents (A)
• aMact ? An
• A set of action languages (Multl)

7
Multi-agent Systems (2/6)
A1
Collaborate
A2
Coordinate
A3
Cooperate
An
8
Multi-Agent Systems Communication Structure
(Shared Memory) (3/6)
Blackboard
Found Survivor (4, 3)
A1
A4
A8
A2
9
Multi-Agent Systems Communication Structure
(Shared Memory) (4/6)

• Advantages
• Easily extends a MAS
• Communication interface is defined
• Disadvantages
• Control
• Access Rights
• Concurrency
• Inform Agents
• Maybe by a modification to the Observer Pattern

10
Multi-Agent Systems Communication Structure
(Message Passing) (5/6)
CommunicationChannel
CommunicationChannel
DirectoryService
Host B
Host B
Host A
Yellowish Agent, please
11
Multi-Agent Systems Communication Structure
(Message Passing) (6/6)

• Advantages
• Several types of communications can be achieved,
  e.g. point-to-point or multicast or broadcast
• Disadvantages
• If no directory service exists
• A new agent has to register with other agents
• Need to define a communication language

12
NASA Project!
Hey!Can you explore(2, 4)
For sure!
13
Standards
14
Standards
FIPA Foundation forIntelligent Physical Ag.
Agents
Agents endpoints
Inter-platformcommunication
European Agentcities
Intra-platformcommunication
15
Communication Languages
16
Agent Communication Languages

• Based on speech act theory
• FIPA ACL
• KQML
• Conversation Protocols
• Social Commitments

Not a language.They are a paradigm
17
Rob Kremer 2004Using a Performative Subsumption
Lattice to Support Commitment-basedConversations
18
Another Example...
19
Collaborative Agent Software Architecture (CASA)
20
Components and Functionality
Acts as the representative of an area
coordinates between agents and yellow pages, may
activate a dormant agent on request
Area
A place where agents execute
Local Area Coordinator
Yellow Pages
Other Agent
Usually registers itself with the LAC can
participate in cooperation domains to offer of
receive request
Allows look up of agents by service
A place where agents communicate. Also a
repository of data associated with the
cooperation domain
Cooperation Domain
21
Concluding Remarks
22
Concluding Remarks (1/3)

• Agent is an autonomous process modeled with some
  (or all) of the following properties
• Social, reactive, proactive, knowledge-based,
  rational, mobile
• Multi-Agent system describes several agents that
  positively interact with each other
• CCC 3 Cs Collaborate, Coordinate, Cooperate
• Main goal Achieve Synergy

23
Concluding Remarks (2/3)

• Agents communicate via
• Shared memory
• Message passing
• Standards set by FIPA
• Agent platform contains
• Agents, agents endpoints, directory services,
  managers, intra-platform communication channel,
  inter-platform communication channel (ACC)

Not mutually exclusive
24
Concluding Remarks (3/3)

• Agents need to communicate understand each
  other!
• Need for an Agent Communication Language (ACL)
  based on speech act theory
• FIPA-ACL KQML are examples of ACLs
• KQML can also be used as a message wrapper
• CASA is a multi-agent platform developed at the
  University of Calgary

25
Questions?