AI: Planning System

Russel and Norvig, Chapter 11

L. Manevitz, Slide Presentasi,
http://www.cs.uiowa.edu
 Overview
The blocks world problem

Initial State Goal

• Bagaimana mendapatkan goal dari initial state ?

 Overview Planning
• Planning adalah metode yang telah dipikirkan
secara detail sebelum menyelesaikan suatu
pekerjaan.
Misalkan: ide atau metode untuk mengalahkan
tim lawan dalam pertandingan sepakbola.
• Planning is action or process of making plans for
something.
• Planning penting untuk solusi yang tidak dapat
dibalik kembali (undo).
 Overview Planning (2)
• Planning adalah kasus khusus dari searching.
• Pada searching, pelacakan dilakukan dari initial state
ke goal state dengan mengikuti path yang ada. Path
yang merupakan solusi adalah sebuah plan langkah
dari initial ke goal state.
• Sedangkan pada planning, tidaklah harus masalah itu
dipecahkan dari initial ke goal state, tetapi dapat juga
dengan memecahkan subgoal.
 Metode Planning
STRIPS (STanford Research Institute Planning System-1971 by
Fikes and Nilsson)
• Goal Stack Planning
• Constraint Posting
• NOAH
• etc
 Komponen planning system
• Memilih rule terbaik berdasarkan informasi
heuristik yang tersedia.
• Mengimplementasikan rule yang dipilih dan
menghitung masalah yang muncul dari
implementasi ini.
• Mendeteksi ketika solusinya tidak ditemukan.
• Mendeteksi jalan buntu(dead end) sehingga bisa
menghindarinya.
• Mendeteksi ketika hampir ditemukan solusi dan
menggunakan metode khusus untuk benar-benar
menemukan solusi.
 Memilih Rule
• Pertama, mendeteksi/mengumpulkan
perbedaan antara goal dan initial dan
kemudian mengidentifikasi rule yang relevan
untuk mengurangi perbedaan ini.
• Jika terdapat beberapa rule yang dapat dipilih,
maka gunakan informasi heuristik untuk
memilih diantara mereka.
 Implementasi Rule
• Setiap rule yang akan diimplementasikan akan
memberikan kondisi yang baru/perubahan.
• Kita harus bisa mengidentifikasi
- kondisi yang harus dipenuhi untuk
menjalankan rule tertentu.
- Akibat dari menjalankan suatu rule tertentu.
 Mendeteksi solusi
• Planning sistem dikatakan menemukan solusi
jika telah menemukan sequence operasi dari
initial state ke goal state.
 Mendeteksi dead ends (jalan buntu)
• Jika perubahan yang dihasilkan setelah iterasi
cukup lama, tidak signifikan untuk mengurangi
perbedaan antara initial dgn goal.
• Jika kita menggunakan metode
forwardchaining: operasi menuju suatu state
yang tidak dapat dicapai oleh operasi dari goal
state.
• Jika metode backward: sebaliknya.
 Teknik mendapatkan solusi
• Jika masalah tersebut dapat
didekomposisi(dipecah-pecah) maka solusi
tiap subbagian telah ditemukan maka solusi
masalah merupakan gabungannya.
 Metode Goal Stack Planning

• Metode ini menggunakan satu stack yang

mengandung goal dan operator yang diusulkan
untuk mencapai goal.
• Adanya set operator yang mendeskripsikan daftar
PRECONDITION, ADD, dan DELETE.
• Jika goal lebih dari satu, metode ini akan
memecahkan goal satu persatu.
• Solusi yang diperoleh yaitu sequence operator
dari goal pertama kemudian diikuti oleh
sequence operator goal kedua dan seterusnya.
 Goal Stack Planning
• Setiap langkah adalah untuk menemukan goal
pada stack teratas.
• Kemudian diikuti oleh goal dibawahnya
sampai semua goal telah dipenuhi.
• Jika terdapat komponen goal yang tidak
dipenuhi maka bagian yang salah dari goal
dikembalikan ke dalam stack dan proses
dilanjutkan kembali.
 Studi kasus
Dunia Balok (Blocks World)

Initial State Goal

Langkah:
1. Mendefinisikan Operator apa saja.
2. Spesifikasi/Deskripsi Operator (PRECONDITION,
ADD dan DELETE).
3. Pendefinisikan Initial dan Goal state.
3. Membuat Stack
4. Implementasi Operator dan menemukan goal
 The Blocks World
• Operators:
– UNSTACK(A,B) – Pick up block A from its current
position on block B. The arm must be empty and
block A must have no blocks on top of it.
– STACK(A,B) – Place block A on block B. The arm
must already be holding A and the surface of B
must be clear.

All rights reserved © L. Manevitz Lecture 6 16

 The Blocks World cont.
– PICKUP(A) – Pick up block A from the table and
hold it. Te arm must be empty and there must be
nothing on top of block A.
– PUTDOWN(A) – Put block A down on the table.
The arm must have been holding block A.
 The Blocks World cont.
• Predicates: (Kondisi)
– ON(A,B) – Block A is on block B.
– ONTABLE(A) – Block A is on the table.
– CLEAR(A) – There is nothing on top of
block A.
– HOLDING(A) – The arm is holding block
A.
– ARMEMPTY – The arm is holding nothing.
 The Blocks World cont.
• Inference rules:
– [ x : HOLDING(x)] ARMEMPTY

– x : ONTABLE(x) y : ON(x,y)

– x:[ y : ON(y,x)] CLEAR(x)

A Simple Blocks World Description

ON(A,B)
ONTABLE(B)
CLEAR(A)
 Spesifikasi Operator
• PRECONDITION: daftar predicate yang harus
benar sebelum suatu operator dipilih.
• ADD: daftar predicate baru (bernilai true)
setelah operator diimplementasikan.
• DELETE: daftar operator lama yang ketika
operator yang diimplementasikan salah.
 Spesifikasi Operator Blocks World
STACK(A, B): PICKUP(A):
P: CLEAR(B)  HOLDING(A) P: CLEAR(A)  ONTABLE(A) 
ARMEMPTY
D: CLEAR(B)  HOLDING(A)
D: ONTABLE(A)  ARMEMPTY
A: ARMEMPTY  ON(A, B)
A: HOLDING(A)
UNSTACK(A, B):
PUTDOWN(A):
P: ON(A, B)  CLEAR(A) 
ARMEMPTY P: HOLDING(A)
D: ON(A, B)  ARMEMPTY D: HOLDING(A)
A: HOLDING(A)  CLEAR(B) A: ONTABLE(A)  ARMEMPTY
 Pendefinisian State

Initial State Goal

start: ON(B, A)  goal: ON(C, A) 

ONTABLE(A)  ON(B, D) 
ONTABLE(C)  ONTABLE(A) 
ONTABLE(D)  ONTABLE(D) 
ARMEMPTY ARMEMPTY

ON(A,B): Balok A menempel diatas balok B

ONTABLE(A): Balok A berada dipermukaan meja
CLEAR(A): Tidak ada balok yg sedang menempel di A
ARMEMPTY: Robot tidak sedang memegang balok
 Membuat Stack

Stack Database

Goals
Current situation

Operators to
satisfy
+ Specification of
Operators/Actions
the Goals
 Implementasi GSP
Kondisi yang mungkin terjadi:
a). Kondisi 1: jika slot berisi kondisi yang sudah
memenuhi current state, tetapi slot ini
terletak diatas operator maka, slot diambil
dari stack dan pemeriksaan dilanjutkan pada
slot berikutnya.
b). Kondisi 2: Operator yang memenuhi suatu
kondisi, dimasukkan ke dalam stack.
c). Kondisi 3: jika slot yang diperiksa adalah slot
paling dasar, maka uji kesamaan antara
current state dan goal state. Jika sama berarti
solusi tercapai.
 Goal Stack Planning
1) Initial goal stack:

ON(C,A)Λ ON(B,D)Λ ONTABLE(A)Λ ONTABLE(D)

Initial State Goal 28

 Goal Stack Planning cont.
2) Choose to work on ON(C,A) before ON(B,D):

ON(C,A)
ON(B,D)
ON(C,A)Λ ON(B,D)Λ OTAD

Initial State Goal 29

 Goal Stack Planning cont.
3) Achieve ON(C,A) with STACK(C,A):

ON(C,A)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD

Initial State Goal 30

 Goal Stack Planning cont.
4) Add STACK’s preconditions:

CLEAR(A)
HOLDING(C)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD

31
Initial State Goal
 Goal Stack Planning cont.
5) Achieve CLEAR(A) with UNSTACK(B,A):

ON(B,A)
CLEAR(B)
ARMEMPTY
ON(B,A) CLEAR(B) ARMEMPTY
UNSTACK(B,A)
CLEAR(A)
HOLDING(C)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD

Initial State Goal 32

 Goal Stack Planning cont.
6) Pop satisfied predicates:

ON(B,A)
CLEAR(B)
ARMEMPTY
ON(B,A) CLEAR(B) ARMEMPTY
UNSTACK(B,A)
HOLDING(C)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD

Initial State Goal 33

 Goal Stack Planning cont.
7) Achieve HOLDING(C) with UNSTACK(C,x):

ON(C,x)
CLEAR(C)
ARMEMPTY
ON(C,x) CLEAR(C) ARMEMPTY
UNSTACK(C,x)
HOLDING(C)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD
Initial State Goal
34
 Goal Stack Planning cont.
8) Achieve ON(C,x) by STACK(C,x):
CLEAR(x)
HOLDING(C)
CLEAR(x) HOLDING(C)
ON(C,x)
STACK(C,x)
CLEAR(C)
ARMEMPTY
ON(C,x) CLEAR(C) ARMEMPTY
UNSTACK(C,x)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD
Initial State Goal
35
 Goal Stack Planning cont.
9) Terminate path because HOLDING(C) is duplicated.
CLEAR(x)
HOLDING(C)
CLEAR(x) HOLDING(C)
STACK(C,x)
CLEAR(C)
ARMEMPTY
ON(C,x) CLEAR(C) ARMEMPTY
UNSTACK(C,x)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD
Initial State Goal 36
 Goal Stack Planning cont.
10) Achieve HOLDING(C) with PICKUP, not UNSTACK:

ONTABLE(C)
CLEAR(C)
ARMEMPTY
ONTABLE(C) CLEAR(C) ARMEMPTY
PICKUP(C)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD

37
Initial State Goal
 Goal Stack Planning cont.
11) Pop ONTABLE(C) and CLEAR(C), and achieve
ARMEMPTY by STACK(B,D):

CLEAR(D)
ONTABLE(C)
HOLDING(B)
CLEAR(C) HOLDING(B)
CLEAR(D)
ARMEMPTY
STACK(B,D)
ONTABLE(C) CLEAR(C) ARMEMPTY
PICKUP(C)
CLEAR(A) HOLDING(C)
STACK(C,A)
ON(B,D)
ON(C,A) ON(B,D) OTAD

38
Initial State Goal
 Goal Stack Planning cont.
12) Pop entire stack, and return plan:
i. UNSTACK(B,A).
ii. STACK(B,D).
iii. PICKUP(C).
iv. STACK(C,A).

Initial State Goal

39
 Algorithma GSP
1. Tempatkan seluruh kondisi goal-state pada slot stack paling bawah.
2. Masukkan setiap kondisi goal state yang belum tercapai ke dalam sebuah
slot stack.
3. Loop
Keluarkan kondisi yang sudang dicapai dari dalam stack.
Ganti kondisi yang belum tercapai dengan operator yang sesuai.
Pindahkan operator yang bisa diaplikasikan kedalam rencana
penyelesaian.
Cek apakah current-state == goal state.
if current state == goal state then
sukses
end
end
Contoh lain
Solusi
 The Problem Solving System NOAH
• NOAH plans by developing a hierarchy of subgoals.
• The procedural net contains several levels of
representation of a plan, each level more detailed
than the previous one.
• For example: the node representing the abstract goal
make coffee may be expanded to : grind coffee, boil
water, put the coffee in a filter, pour the water
through it.

All rights reserved © L. Manevitz Lecture 6 48

All rights reserved © L. Manevitz Lecture 6 49
 NOAH cont.
• We’ll use NOAH to solve the blocks problem.
• The operators used in this example are slightly
different from those we have been using .
• STACK – will put any object on any other (including
the table), provided that both objects are clear. It
includes the picking up of the object to be moved.

All rights reserved © L. Manevitz Lecture 6 50

 The Blocks Problem
• A reminder of the problem :
A

C B

A B C

Start: ON(C,A) Goal: ON(A,B)

ONTABLE(A) ON(B,C)
ONTABLE(B)
ARMEMPTY

All rights reserved © L. Manevitz Lecture 6 51

 NOAH Solution
• The initial state of the problem solver:

(ON(A,B) ON(B,C)) Level 1

All rights reserved © L. Manevitz Lecture 6 52

 NOAH Solution cont.
• The first thing that it does is to divide the
problem into two subproblems:
ON(A,B)

S J Level 2

ON(B,C)

Split Join

All rights reserved © L. Manevitz Lecture 6 53

 NOAH Solution cont.
• At the third level the preconditions of STACK are
considered – the two blocks involved must be clear.

Level 3: before criticism

1 CLEAR(A)
S J 3 STACK(A,B)
2 CLEAR(B)
S J
4 CLEAR(B)
S J 6 STACK(B,C)
5 CLEAR(C)

All rights reserved © L. Manevitz Lecture 6 54

 NOAH Solution cont.
• Now NOAH employs a set of critics to examine
the plan and detect interactions among the
subplans.

• Each critic is a little program that makes

specific observations about the proposed
plan.

All rights reserved © L. Manevitz Lecture 6 55

 The Resolve Conflicts critic
• The Resolve Conflicts critic – constructs a table
that lists all the literals that are mentioned
more than once in the plan.

CLEAR(B): asserted: node 2 “Clear B”

denied: node 3 “Stack A on B”
asserted: node 4 “Clear B”
CLEAR(C): asserted: node 5 “Clear C”
denied: node 6 “Stack B on C”

All rights reserved © L. Manevitz Lecture 6 56

 The Resolve Conflicts critic cont.
• Constraints on the ordering of operation arise
when a given literal must be true before one
operation can be performed but will be
undone by another.

CLEAR(B): denied: node 3 “Stack A on B”

asserted: node 4 “Clear B”

All rights reserved © L. Manevitz Lecture 6 57

 The Resolve Conflicts critic cont.
• Conclusion:
Since putting A on B could undo the
preconditions for putting B on C, putting B on
C needs to be done first.

All rights reserved © L. Manevitz Lecture 6 58

 NOAH Solution cont.
• The third level after criticism to resolve
conflicts:
Level 3

CLEAR(A)
S J STACK(A,B)
CLEAR(B)
S
CLEAR(B)
S J STACK(B,C)
CLEAR(C)

All rights reserved © L. Manevitz Lecture 6 59

Eliminate Redundant Preconditions
• Can be invoked to eliminate the redundant
specification of subgoals. In this case CLEAR(B)
appears twice and is only denied by the last
step of the plan.

All rights reserved © L. Manevitz Lecture 6 60

 NOAH Solution cont.
• The third level after all criticism:

Level 3
CLEAR(A)

J STACK(A,B)
S

CLEAR(B)
S J STACK(B,C)
CLEAR(C)

All rights reserved © L. Manevitz Lecture 6 61

 NOAH Solution cont.
• To clear A, C must be removed from A. in
order to do that C must be clear :
Level 4: before criticism
CLEAR(C) STACK(C,x)

S J STACK(A,B)

CLEAR(B)
S J STACK(B,C)
CLEAR(C)

All rights reserved © L. Manevitz Lecture 6 62

 The Resolve Conflicts critic
• The critic observes that putting B on C will
make CLEAR(C) false, so everything that
depends on C being clear will have to be done
before B is put on C.

All rights reserved © L. Manevitz Lecture 6 63

 NOAH Solution cont.
• The Resolve Conflicts critic is employed:
CLEAR(C) STACK(C,x)

J STACK(B,C) STACK(A,B)
CLEAR(B)
S
CLEAR(C) Level 4: after criticism to
resolve conflicts

All rights reserved © L. Manevitz Lecture 6 64

Eliminate Redundant Preconditions
• It notices that CLEAR(C) is required twice.
• Since putting C somewhere must occur before
putting B on C, and both require C being clear
then we know that when we get to putting B
on C, C will be clear.
• CLEAR(C) can be eliminated from the lower
path.

All rights reserved © L. Manevitz Lecture 6 65

 NOAH Solution cont.
• The Eliminate Redundant Preconditions critic
is called:
Level 4: after all criticism

CLEAR(C) STACK(C,x)

CLEAR(B) J STACK(B,C) STACK(A,B)

All rights reserved © L. Manevitz Lecture 6 66

 NOAH Solution cont.
• The system observes that the remaining goals
, CLEAR(C) and CLEAR(B), are true in the initial
state.
• Therefore the final plan is:
Final plan

STACK(C,x) STACK(B,C) STACK(A,B)

All rights reserved © L. Manevitz Lecture 6 67