Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial...
-
Upload
timothy-davidson -
Category
Documents
-
view
213 -
download
0
Transcript of Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial...
![Page 1: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/1.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Lecture 20 of 42
Introduction to Classical Planning:STRIPS & Partial-Order Planning (POP)
William H. Hsu
Department of Computing and Information Sciences, KSU
KSOL course page: http://snipurl.com/v9v3
Course web site: http://www.kddresearch.org/Courses/CIS730
Instructor home page: http://www.cis.ksu.edu/~bhsu
Reading for Next Class:
Section 11.3, p. 387 – 394, Russell & Norvig 2nd edition
Partial plan: http://en.wikipedia.org/wiki/Partial_plan
![Page 2: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/2.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Lecture Outline
Reading for Next Class: Section 11.3 (p. 387 – 394), R&N 2e
Last Class: Knowledge Representation Concluded; Midterm Review Inheritance semantics Midterm exam emphasis
Rational intelligent agents: reflex, reflex/state, goals, preferences
Search: heuristic, constraint, game tree
Knowledge representation and inference: logic, resolution; FC/BC, LSATC
Today: Classical Planning, Sections 11.1 – 11.2 (p. 375 – 386), R&N 2e
Planning problem defined
Initial conditions
Actions: preconditions, postconditions
Goal conditions / goal test Limitations of situation calculus and FOL STRIPS operators: represent actions with preconditions, ADD/DELETE lists
Coming Week: Midterm; More Classical and Robust Planning
![Page 3: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/3.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Planning in Situation Calculus
© 2004 S. Russell & P. Norvig. Reused with permission.
![Page 4: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/4.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Making Plans using FOL:Review
© 2004 S. Russell & P. Norvig. Reused with permission.
![Page 5: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/5.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Making Plans – Better Way:Review
© 2004 S. Russell & P. Norvig. Reused with permission.
![Page 6: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/6.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
STRIPS Operators
Adapted from materials © 2003 – 2004 S. Russell & P. Norvig. Reused with permission.
![Page 7: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/7.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
State Space versus Plan Space
© 2004 S. Russell & P. Norvig. Reused with permission.
![Page 8: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/8.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Air Cargo Transport ProblemSTRIPS Specification
© 2003 S. Russell & P. Norvig. Reused with permission.
Figure 11.2p. 380 R&N 2e
![Page 9: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/9.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
STRIPS and Its Limitations:Need for Richer Planning Language
What STRIPS Can Represent States Goals Actions (using action schema)
Preconditions: must be true before action can be applied
Effects: asserted afterwards
Real STRIPS: ADD, DELETE Lists for Operators STRIPS Assumption
Representational frame problem solution Default is that conditions remain unchanged unless mentioned in effect
What STRIPS Cannot Represent Negated preconditions Inequality constraints
Richer Planning Language: Action Description Language (ADL)
![Page 10: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/10.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
STRIPS vs.Action Description Language (ADL)
© 2003 S. Russell & P. Norvig. Reused with permission.
Figure 11.1p. 379 R&N 2e
![Page 11: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/11.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Simple Spare Tire Problem [1]:Illustrated Example
© 2004 S. Russell & P. Norvig. Reused with permission.
![Page 12: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/12.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Simple Spare Tire Problem [2]:ADL Specification
© 2003 S. Russell & P. Norvig. Reused with permission.
Figure 11.3p. 381 R&N 2e
![Page 13: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/13.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Blocks World:Three-Block Tower Problem
© 2003 S. Russell & P. Norvig. Reused with permission.
Figure 11.4p. 383 R&N 2e
![Page 14: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/14.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Forward (Progression) vs. Backward (Regression) State Space Search
© 2003 S. Russell & P. Norvig. Reused with permission.
Figure 11.5p. 383 R&N 2e
![Page 15: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/15.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Failure of Non-Interleaved Planning:Sussman Anomaly in Blocks World
Adapted from slides © 2004 S. Russell & P. Norvig. Reused with permission.
![Page 16: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/16.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Search versus Planning:State Space Search
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
![Page 17: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/17.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Partial Order Planning (POP) [1]:Total Order Plans & Interleavings
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
Figure 11.6p. 389 R&N 2e
![Page 18: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/18.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Partial Order Planning (POP) [2]:Definition – Complete Plans
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
![Page 19: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/19.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
POP Algorithm [1]:Top-Level Functions
![Page 20: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/20.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
POP Algorithm [2]:Lower-Level Functions & Properties
![Page 21: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/21.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Clobbering andPromotion / Demotion
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
![Page 22: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/22.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Preview:How Things Go Wrong in Planning
Based on slide © 2004 S. Russell & P. Norvig. Reused with permission.
![Page 23: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/23.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Terminology
Classical Planning – STRIPS and ADL Planning problem defined
Initial conditions
Actions: preconditions, effects (postconditions)
Goal conditions / goal test STRIPS operators: action specifications ADL operators: allow negated preconditions, unequality
Partial-Order Planning Represent multiple possible interleavings Keep track of which ones are achievable Complete plans
Every precondition achieved,
No clobberings by possibly intervening steps
Sussman Anomaly Contains threat that needs to be resolved to get to goal Illustrates need for partial-order planning, promotion / demotion
![Page 24: Computing & Information Sciences Kansas State University Lecture 20 of 42 CIS 530 / 730 Artificial Intelligence Lecture 20 of 42 Introduction to Classical.](https://reader035.fdocuments.us/reader035/viewer/2022062713/56649ced5503460f949bb117/html5/thumbnails/24.jpg)
Computing & Information SciencesKansas State University
Lecture 20 of 42CIS 530 / 730Artificial Intelligence
Summary Points
Last Class: Knowledge Representation Concluded; Midterm Review Inheritance semantics Midterm emphasis: intelligent agents, search, KR, resolution/unification
Today: Classical Planning – STRIPS and ADL Planning problem defined
Initial conditions
Actions: preconditions, postconditions
Goal conditions / goal test Limitations of situation calculus and FOL STRIPS operators ADL operators: allow negated preconditions, unequality
Next Time (After Exam): More Classical and Robust Planning Hierarchical abstraction planning (ABSTRIPS) Robust planning: sensorless, conditional, monitoring/replanning, continual
Coming Week: Midterm; Planning Continued