Basic BDI Agent - PowerPoint PPT Presentation

1 / 22
About This Presentation
Title:

Basic BDI Agent

Description:

Desires represent states of affairs (soa) that the agent wants ... Once there is commitment one might blindly trudge ahead or. Revise plans when a plan fails or ... – PowerPoint PPT presentation

Number of Views:145
Avg rating:3.0/5.0
Slides: 23
Provided by: danroc
Category:
Tags: bdi | agent | basic | trudge

less

Transcript and Presenter's Notes

Title: Basic BDI Agent


1
Basic BDI Agent
  • Dan Rochowiak
  • drochowi_at_cs.uah.edu

2
The basics
Desired Agent Properties Autonomy Proactive Reac
tive Social
Agent
Sensor Input
Action Output
Environment
3
Intuitions
  • Beliefs (Bel) represent the information the agent
    has about the world.
  • Desires represent states of affairs (soa) that
    the agent wants to be true
  • Intentions (Int) represent deires that the agent
    has committed to achieving.

4
Formalism example
  • (Int i soa) gt (Bel i soa)
  • If the agent (i) intends the state of affairs
    (soa), then it is not the case that the agent (i)
    believes that the state of affairs (soa) obtains.

5
Practical reasoning
  • Theoretical reasoning is about belief.
  • In theoretical reasoning the concern is for the
    transfer of truth or its surrogate
  • Practical reasoning is about action not belief.
  • In practical reasoning the concern is for the
    production of an action

6
Practical reasoning
  • Practical reasoning involves deliberation and
    planning (means-ends)
  • Deliberation focuses on deciding what soa we want
    to achieve
  • Means-ends reasoning (planning) focuses on
    deciding how to achieve the specified soa.
  • All practical reasoning understood as computation
    is resource bound

7
Intentions
  • Intentions lead to action and drive means-end
    reasoning (pro-active).
  • Intentions endure over time (persistence)
  • Intentions constrain future deliberation.
  • Intentions refer to future soa.

8
Two cases
  • Having the intention to bring about soa while
    believing that you will not bring about soa is
    called intention-belief inconsistency and is held
    to be irrational
  • Having the intention to bring about soa while not
    believing that soa will be the case is called
    intention-belief incompleteness and is rational

9
The other cases
  • Having the intention to bring about soa while
    believing that soa is the case is
  • Having the intention to bring about soa while not
    believing that soa is the case is
  • Having the intention to bring about soa while
    believing that soa will be the case is

10
Basic agent control
  • While true
  • Observe environment
  • Update internal world model
  • Deliberate about intention
  • Get plan for intention
  • Execute plan
  • End while

11
Plans
  • Preconditions define the circumstances under
    which the plan is applicable
  • Postconditions define the soa the plan achieves
  • Body is the sequence of actions
  • Let pl be a plan and Plans be the set of all plans

12
Plans
  • If pl is a plan, then pre(pl), post(pl) and
    body(pl) are the precondition, the postcondition
    and the body of the plan.
  • If pl is a plan, then empty(pl) is an empty
    sequence.
  • If pl is a plan, then execute(pl) is a procedure
    that executes the plan without stopping

13
Plans
  • If pl is a plan, then head(pl) is the plan with
    the first action of body(pl)
  • If pl is a plan, then tail(pl) is the plan with
    all but the first action of body(pl)
  • If pl is in Plans, i is in Int and b is in Bel,
    then sound(pl,i, b) means that pl is a sound plan
    for achieving i.

14
As the world changes
  • The function brf(b,p) is the belief revision
    function that changes Beliefs given a new percept
    p.
  • The function deliberate(b) returns an intention
    from Beliefs
  • The function plan(b,i) returns a plan from the
    current Beliefs and Intentions

15
New control
  • Let B be Beliefs
  • While true
  • Get next p (percept)
  • Let B be brf(B,p)
  • Let I be deliberate(B)
  • Let pl be plan(B,I)
  • execute(pl)
  • End while

16
Deliberation
  • Deliberation has two parts
  • Generate a set of possible alternatives
  • The function options(B,I) takes Beliefs and
    Intentions and generates Desires
  • Select between competing alternatives and commit
    to achievement
  • The function filter(B,D,I) takes Beliefs,
    Desires, and Intentions and returns a new
    Intentions

17
Revised control
  • Let B be Beliefs
  • Let I be Intentions
  • While true
  • Get next p
  • Let B be brf(B,p)
  • Let D be options(B,I)
  • Let I be filter(B,I)
  • Let pl be plan(B,I)
  • execute(pl)
  • End while

18
Commitment
  • Once there is commitment one might blindly trudge
    ahead or
  • Revise plans when a plan fails or
  • Drop impossible or successful intentions

19
The final general version
  • Let B be Beliefs
  • Let I be Intentions
  • While true

20
The final general version
  • Get next p
  • Let B be brf(B,p)
  • Let D be options(B,I)
  • Let I be filter(B,D,I)
  • Let pl be plan(B,I)

21
The final general version
  • While not (empty(pl) or succeeded(I,B) or
    impossible(I,B))
  • Let A be head(pl)
  • execute(A)
  • Let pl be tail(pl)

22
The final general version
  • Get next p
  • Let B be brf(B,p)
  • If not sound(pl,I,B) then
  • Let pl be plan(B,I)
  • End if
  • End while
  • End while
Write a Comment
User Comments (0)
About PowerShow.com