Artificial Intelligence Russell - PowerPoint PPT Presentation

1 / 14
About This Presentation
Title:

Artificial Intelligence Russell

Description:

not able to directly feed into agents effectors. Low Level Abstraction ... Negated & Disjunctive Goals. Negated Literals. Unification Function. p p ... – PowerPoint PPT presentation

Number of Views:113
Avg rating:3.0/5.0
Slides: 15
Provided by: ssba5
Category:

less

Transcript and Presenter's Notes

Title: Artificial Intelligence Russell


1
Artificial IntelligenceRussell
NorvigChapter 12
  • By
  • Sukhvinder Bahra

2
Practical Planners
  • OPTIMUM-AIV
  • Generate Plans Monitors
  • Replan
  • Automatically generate plans
  • Input Output Representations
  • - Must be expressive easy to understand

3
STRIPS inefficiency
  • Hierarchical Plans
  • Complex Conditions
  • Time
  • Resources

4
Hierarchical Decomposition
  • High Level Abstraction
  • - not able to directly feed into agents
    effectors.
  • Low Level Abstraction
  • - too much time and steps.
  • Abstract Operator
  • Primitive Operator

5
Extending the Language
  • Partition Operators
  • - Primitive Operators
  • - Non-Primitive Operators
  • Add set of Decomposition Methods.

6
Analysis of Hierarchical Decomposition
  • Subroutines are good
  • - can be combined to create large plans.
  • Abstract solution
  • Two solutions
  • - downward solution
  • - upward solution

7
  • A simplified model
  • Assume that there is one solution with n
    primitive steps.
  • Time to choose right steps.

8
  • (see figure12.5)
  • Non-Hierarchical Planner
  • - worst case O(bn)
  • Hierarchical Planner
  • - search decompositions to abstract solution

9
Decomposition v Approximation
  • Hierarchical Planning
  • - Hierarchical Decomposition
  • abstract, non-primitive operator
  • decomposed into steps.
  • - Abstraction Hierarchies
  • single operator planned at different
  • levels of abstraction

10
Negated Disjunctive Goals
  • Negated Literals
  • Unification Function
  • p p
  • Initial State - dont specify false conditions
  • Disjunctive PreConditions - p v q
  • Disjunctive Effects - hard to incorporate

11
Universal Quantification
  • Extend for Universally Quantified PreCondition.
  • eg, Clear(b)
  • Extend for Universally Quantified Effects.
  • eg, Carry(bag,x,y)

12
Condition Effects
  • Operators have different meanings
  • - Blocks world
  • Initial state On(a,b)
  • Goal state Clear(b)
  • Single Operator
  • - e.g, Move(b,x,y)
  • effect when condition

13
Measures of Planning
  • Measures can be logical terms.
  • Measures that change over time.
  • Planners need declarations up-front.
  • Resource Measures
  • e.g, Cash, GasLevel
  • Effect operator
  • left-handside is a measure.
  • right-handside is expression involving
    measure.

14
Temporal Constraints
  • Time is a resource.
  • Initial-state has a start-time for plan.
  • Add time to time for each operation.
  • e.g, 10 seconds to pump gallon of gas.
  • 3 minutes for rest of filling up.
Write a Comment
User Comments (0)
About PowerShow.com