Distributed Reasoning: Agent Systems

1
Distributed ReasoningAgent Systems

• Bob McKay
• School of Computer Science and Engineering
• College of Engineering
• Seoul National University

2
Outline

• Characterisation of agent systems
• Classes of Agent Systems
• BDI Agents
• Distributed Agents

3
Intelligent Agents

• Object systems
• provide encapsulation of methods and data
• Provide decentralisation of code
• through message passing
• still need to know where to send messages
• need to know where capabilities reside in the
  system
• Agent systems
• an amalgam of ideas from
• blackboard systems
• object systems

4
Characteristics of Intelligent Agents

• Encapsulated
• As with objects
• Autonomous
• The individual agents are able to run
  independently
• Though not necessarily usefully
• Distributed
• Most useful agent architectures are distributed
  (usually internet capable)
• Personalisable
• learning, reasoning from previous cases
• Goal Driven
• The agents actions are guided by goals they are
  aiming to accomplish
• either individually or in cooperation

5
Characteristics of Intelligent Agents

• Discourse
• Need a flexible and extensible language of
  communication
• Capable of representing many aspects of the real
  World
• Negotiation about Expectations
• Rather than direct message passing
• Risk and Trust
• An individual agent wont necessarily be able to
  guarantee
• The ability to carry out a particular task
• That the answers obtained are correct
• Graceful Degradation
• If the system is unable to accomplish a
  particular goal
• Failure is graceful
• If possible, failure is partial

6
Types of Intelligent Agents

• Logic-Based Agents
• Reactive Agents
• Belief-Desire-Intention (BDI) Agents

7
Logic-Based Agents

• An agent is a logic-based KR system
• Uses a symbolic logic model of the world
• Decisions using logical deduction
• Problems
• Frame problem etc
• standard problems of logic representation
• Updating the logic model in real time

8
Reactive Agents

• Dont maintain an explicit representation of the
  World
• Based in the ideas of situated cognition
• The world supplies all the necessary complexity
• Dont require logical models for this
• Drive toward a goal
• Avoid collisions
• Obey stop signs
• ..
• Each intelligent component is simple
• The World supplies the appropriate
  inter-relationship

9
BDI Agents

• Beliefs
• The information an agent has about its
  surroundings
• The agents model of the current state
• Desires
• The things that an agent would like to see
  achieved
• Desired changes of state
• Intentions
• The desires that an agent is attempting to
  achieve
• Not all desires can be achieved
• A hopefully achievable subset
• Based in a modal logic with three modal operators
• Bel
• Des
• Int

10
Beliefs

• A Bel p iff p is entailed in every possible world
  the agent believes it could be in
• Doesnt satisfy Kp ? KKp
• Bounded rationality (limited resources) the
  agent cant necessarily derive all consequences
  of its beliefs

11
Desires

• A Des p iff p holds in all desired worlds
  reachable from the present state
• The agent might not know how to reach the states
  it desires to be in
• An agent can desire to be in conflicting states
• Goals are the subset of the agents desires that
  are achievable and consistent

12
Intentions

• Goals the agent has committed to achieve
• May nevertheless not be achievable
• The agent may be mistaken
• May not achieve it even if achievable
• Agent may proceed along an unintended path

13
Summary

• Characterisation of agent systems
• Classes of Agent Systems
• BDI Agents
• Distributed Agents