Multiagent Systems and Societies of Agents

Authors: Michael N. Huhns and Larry M. Stephens

Speaker: Lin Xu (part I) and Shabbir Syed (part II)

CSCE 976, April 3rd 2002

Outline1. Introduction2. Agent communications

Coordination Dimensions of meaning Message types Communication levels Speech acts, KQML, KIF, Ontology, other

3. Agent interaction protocols4. Societies of agents5. Conclusions

Introduction

How to analyze, describe, and design environments in which agents can operate effectively and interact with each other productively.

Communication protocols [Xu Lin] Enable agents to exchange and understand messages

Interaction protocols [Shabbir Syed] Enable agents to have conversations, which are

structured exchanges of messages

Communication protocols

Enable agents to exchange and understand messages

The messages can be exchanged between two agents: Propose a course of action Accept a course of action Reject a course of action Retract a course of action Disagree with a proposed course of action Counter-propose a course of action

Interaction protocols Enable agents to have

conversations, which are structured exchanges of messages

Negotiation can occur between Agent1 and Agent2

Agent1 proposes a course of action to Agent2

Agent2 evaluates the proposal and Sends acceptance to Agent1 or Sends counterproposal to Agent1 or Sends disagreement to Agent1 or Sends rejection to Agent1

Motivation

Centralized solutions are generally more efficient, why should we interested in distribution system?

Easier to understand and easier to develop, when the problem being solved is itself distributed.

Lead to computational algorithms that might not have been discovered with a centralized approach.

A centralized approach is impossible. Respect real conditions: privacy of agents, distribution

Characteristics of Multiagent Environments

Provide an infrastructure specifying communication and interaction protocols

Typically open and have no centralized design

Contain agents that are autonomous and distributed, and may be self-interested or cooperative

Agent Communications

An agent is an active object with the ability to perceive, reason, and act

An agent has explicitly represented knowledge and a mechanism for operating on or drawing inferences from its knowledge

An agent has the ability to communicate (receiving messages and sending messages)

Communications

1. Coordination2. Dimensions of meaning3. Message types4. Communication levels 5. Examples:

a. Speech actsb. KQMLc. KIFd. Ontologies e. Other…

Coordination

A property of a system of agents performing some activity in a shared environment

Avoid extraneous activity by reducing resource contention

Avoid livelock and deadlock Maintain applicable safety conditions

Cooperation: Among non-antagonistic agentsNegotiation: Among competitive/self-interested

agents

Ways for coordinating behavior and activities among agents

How well a system behaves as a unit?

How it can maintain global coherence without explicit global control Be able to determine on their own goals they

share with other agents Determine common task Avoid unnecessary conflicts Pool knowledge and evidence

Some organization among the agents is needed

Dimensions of meaning

Three aspects to the formal study of communication:

Syntax: how the symbols of communication are structured

Semantics: what the symbol denote Pragmatics: how the symbol are interpreted

Meaning is a combination of semantics and pragmatics

Different dimensions of meaning associated with communication

Descriptive vs. Prescriptive Personal vs. Conventional meaning Subjective vs. Objective meaning Speaker’s vs. Hearers’s vs. Society’s

Perspective Semantics vs. Pragmatics Contextually Coverage Identity Cardinality

Message types

Two basic message types: assertions and queries

Basic agent: accept assertions Passive role (answer questions): accept a

query, send a reply, accept information Active role: issue queries, make assertions,

accept assertion Peer: assume both active and passive role in

dialog

Message types

Two basic message types: assertions and queries

Dialogue vs. FunctionActive MasterPassive SlaveBoth Both

Communication levels

Communication protocols are typically specified at several levels:

Lowest level: specifies the method of interconnection

Middle level: specifies the format, or syntax, of the information being transferred.

Top level: specifies the meaning, or semantics, of the information.

Communication levels (cont’d)

There are both binary and n-ary communication protocols:

Binary: a single sender and a single receiver N-ary: a single sender and multiple receivers

A protocol is specified by a data structure with 5 fields: Sender Receiver(s) Language in the protocol Encoding and decoding functions Actions to be taken by the receiver(s)

Speech Act (I)

A popular basis for analyzing human communication is speech act theory

Speech act theory views human natural language as actions

Spoken human communication is used as the model for communication among computational agents

Speech Act (II)

A speech act has three aspects: Locution: the physical utterance by the speaker. Illocution: the intended meaning of the

utterance by the speaker. Perlocution: the action that results from the

locution.

Speech act theory helps define the type of message by using the concept of illocutionary force, whichconstraints the semantics of the communication act

itself

Knowledge Query and Manipulation Language (KQML) [Finin 94]

KQML is a protocol for exchanging information and knowledge.

The basic KQML

Information for understanding the content of the message is includes in the communication itself

(KQML-performative:sender <word>:receiver <word>:language <word>:ontology <word>:content <expression>…)

Syntax is Lisp-like :--)

Nested KQML message

Seven basic categories of KQML

1. Basic query performatives2. Multiresponse query performatives3. Response performatives4. Generic informational performatives5. Generator performatives6. Capability-definition performatives7. Networking performatives

Issues

The sender and receiver must understand the agent communication language

The ontology must be created and be accessible to the agents that are communicating

KQML must operate within a communication infrastructure that allows agents to locate each other

KQML is still a work in progress and its semantics have not been completely defined [1987]

Knowledge Interchange Format (KIF) [Genesereth?]

A logic language proposed as a standard to describe facts in expert systems, database, intelligent agents, etc.

Specifically designed to serve as an “interlingua” or mediator in the translation of other languages

KIF is a prefix version of first order predicate calculus with extensions to support non-monotonic reasoning and definitions. It also can be used to describe procedures.

Ontologies [Fikes et al.]

A specification of objects, concepts, and relationships in an area of interest

The classes and relationships must be represented in the ontology

An agent must represent its knowledge in the vocabulary of a specified ontology

Other communication protocols

Speech Act, KQML, KIF, Ontology in no way preclude other means by which agents can interact, communicate, and be interconnected

Once communication protocols are defined and agreed upon by a set of agents, higher level protocols can be readily implemented

Interaction Protocols

Questions?If not, let’s start the

discussion