1 INTRODUCTION
1

2 INTELLIGENT AGENTS
function TABLE -D RIVEN -AGENT( ¨§¦¤¢
© ¡¥£¡
) returns an action
static: ¨§¦¤¢
© ¡¥£¡
, a sequence, initially empty
¤¨©
¡  
, a table of actions, indexed by percept sequences, initially fully specified
append §¦§¢
© ¡¥£¡ to the end of¨§¦¤¢
© ¡¥£¡
% # \" © ¥
&$!! L OOKUP( , ¡   © ¡ ¥ £ ¡
(§'¨© ¨§¦¤¢ )
return $!¤
# \" ©¥
Figure 2.8
function R EFLEX -VACUUM -AGENT( [ 4!$¨2 $!10)
 3© © # \" © ¥ \"
, ]) returns an action
if 7© £
!§6 5 4!$¨2=  3© © then return 9¥
¤¢3 8
else if # \" © ¥ \"
$!!$0)@ = A then return © D C
'1@ B
# \" © ¥ \"
E $!!$0)@
else if = then return ©G
0H¡ F
Figure 2.10
function S IMPLE -R EFLEX -AGENT( §¦§¢
© ¡¥£¡
) returns an action
static: §(I§£
 ¡ 3
, a set of condition–action rules
P1¨0
% ¡© © I NTERPRET-I NPUT( 02§¢
© ¡¥£¡ )
% ¡ 3
P(I§£ RULE -M ATCH( ,  ¡ 3 ¡ © ©
¤4§£ $Q0 )
%&$!!
# \" © ¥ RULE -ACTION[ ] I§£
¡ 3
return $!¤
# \" ©¥
Figure 2.13
2

3
function R EFLEX -AGENT-W ITH -S TATE( ¨§¦¤¢
© ¡¥£¡ ) returns an action
static: $Q0
¡© ©
, a description of the current world state
§(I§£
 ¡ 3
, a set of condition–action rules
#$!¤
\" ©¥
, the most recent action, initially none
% ¡© ©
P1¨0 §¦§¢ $!¤ $Q0
© ¡¥£¡ # \" ©¥ ¡© ©
U PDATE -S TATE( , , )
% ¡ 3
P(I§£ RULE -M ATCH( ,  ¡ 3 ¡ © ©
¤4§£ $Q0
)
%&$!!
# \" © ¥ RULE -ACTION[ ] I§£
¡ 3
return $!¤
# \" ©¥
Figure 2.16

3 SOLVING PROBLEMS BY
SEARCHING
function S IMPLE -P ROBLEM -S OLVING -AGENT( R§¦¤¢
© ¡¥£¡
) returns an action
inputs: R§¦¤¢
© ¡¥£¡, a percept
static: 2
S¡
, an action sequence, initially empty
$Q0
¡© ©
, some description of the current world state
1)'C
  \"
, a goal, initially null
§(§'¦
T ¡  \" £
, a problem formulation
U PDATE -S TATE( , )
P1¨0
© ¡¥£¡ ¡© ©
% ¡© ©
02§¢ $Q0
if is empty then do 2¤
S¡
F ORMULATE -G OAL( $)C
% \"
) 1¨0
¡© ©
%UT¤¤2R
¡  \"£
F ORMULATE -P ROBLEM(   \" ¡© ©
$)'C $Q0
, )
T§(§'\"¦£
¡ S EARCH( ) V2¤
% S¡
S2¡
F IRST( ) &$!!
% # \" © ¥
R EST( ) S¦¡ W2
% S¡
return $!¤
# \" ©¥
Figure 3.2
function T REE -S EARCH( , T ¡  \" £
§(§'¦ 7 C¡© £©
$Q12!0
) returns a solution, or failure
initialize the search tree using the initial state of §(§'¦
T ¡  \" £
loop do
if there are no candidates for expansion then return failure
choose a leaf node for expansion according to $Q12!0
7 C¡© £©
if the node contains a goal state then return the corresponding solution
else expand the node and add the resulting nodes to the search tree
Figure 3.9
4

Figure 3.17
if then return £ else return
¡0I3 6$¤G q\" 3
sQ©4¤¥ r a¡££ 3 ¥¥ q\"© 3
¢22§I0)\" s¨I¤¥
else if then return © 42¤2£
3¡ 0I3 6$¤G v© I0¤2£
¡ £   uw  3  ¡
if = then true Pr¢22§£I0)\" qs\"¨©I3¤¥
% a¡£ 3 ¥¥ s¨I¤¥ © I0¤2£
q\"© 3  3¡
R ECURSIVE -DLS( © T , , )
6p6  §¡(§'¦ $¤0§¡§0¢30
T  \"£ £ \" ¥¥ RtI0§2£
% © 3  ¡
else for each in E XPAND( , ) do
T¡  \"£
¤¤'¦ ¡')# a \" £$¤2¤§0¢0
\"¡¥¥ 3
else if D EPTH[ ]= then return \"© 3
qsQ4¤¥ © 6pTH  ¡ a \"
')b#
if G OAL -T EST[ ¡ a \"
')# ](S TATE[ ]) then return S OLUTION( )
¡ a \"
')b# T ¡  \" £
§(§'¦
false P¢22§I0)\" ¨I¥
% r a¡££ 3 ¥¥ q\"© 3
function R ECURSIVE -DLS( , , ) returns a solution, or failure/cutoff
© T
6p6  T ¡  \" £ ¡ a \"
§(§'¦ ')b#
return R ECURSIVE -DLS(M AKE -N ODE(I NITIAL -S TATE[
6p6 
© T T ¡  \" £
§(§'¦ T ¡  \" £
§(§'¦ ]), , )
function D EPTH -L IMITED -S EARCH( , ) returns a solution, or failure/cutoff
© T
6pH  T ¡  \" £
¤¤2R
Figure 3.12
return §$¤0§§0¢0
£ \"¡¥¥ 3
add to £ \"¡¥¥ 3
§$¤0§§0¢0 
D EPTH[ ] D EPTH[ ]+1 ')#
¡ a \" % 
, ) $!¤ ')b#
# \" ©¥ ¡ a \" PATH -C OST[ ] PATH -C OST[ ] + S TEP -C OST(
¡ a \"
')# , % 
ACTION[ ] # \" ©¥ 
$!¤i% 
PARENT-N ODE[ ] ¡'a)\"b#h% 
S TATE[ ] © I0¤2g% 
 3¡£
a new N ODE % f
]) do )b#
¡ a \" for each ,
¤¤'¦
T ¡  \" £in S UCCESSOR -F N[ ](S TATE[ e  3¡ # \" ©¥
¦© I0¤2£ $!¤ d
the empty set % £ \"¡¥¥ 3
c§$¤0§§0¢0
) returns a set of nodes T ¡  \" £ X ¡ a \"
¤¤'¦¨)b# function E XPAND(
'I6§0G
¡ C # £ I NSERT-A LL(E XPAND(
¤¤'¦ ')#
T ¡  \" £ ¡ a \" , ), ) `'I6§0G
% ¡ C # £
then return S OLUTION( ) ¡ a \"
')#
if G OAL -T EST[
')b#
¡ a \" ] applied to S TATE[ ] succeeds ¤¤'¦
T ¡  \" £
R EMOVE -F IRST( ) I'H§0G
¡ C # £ % ¡ a \"
P')#
if E MPTY ?( ) then return failure ¡ C # £
I'6§§G
loop do
) 'I6§0G
¡ C # £ I NSERT(M AKE -N ODE(I NITIAL -S TATE[
T ¡  \" £
§(§'¦ ]), % ¡ C # £
`I'H§0G
) returns a solution, or failure ¡ C # £ GX T ¡  \" £
'I6§0HY§(§'¦ function T REE -S EARCH(
5

6 Chapter 3. Solving Problems by Searching
function I TERATIVE -D EEPENING -S EARCH( ¤¤'¦
T ¡  \" £ ) returns a solution, or failure
inputs: §(§'¦
T ¡  \" £
, a problem
for xHa
% D© ¡ 0 to y do
RtI0§2£
% © 3  ¡ D EPTH -L IMITED -S EARCH( ¤¤'¦
T ¡  \" £ , !R'a
D© ¡ )
uv© I0¤2£
w  3¡
if cutoff then return tI0§0£
© 3  ¡
Figure 3.19
function G RAPH -S EARCH( 'I6§0H€¤¤'¦
¡ C # £ GX T ¡  \" £ ) returns a solution, or failure
2$@¥
% a ¡  \" an empty set
% ¡ C # £
`I'H§0G I NSERT(M AKE -N ODE(I NITIAL -S TATE[ ]), ) T ¡  \" £
§(§'¦ ¡ C # £
'I6§0G
loop do
if E MPTY ?( I'6§§G
¡ C # £
) then return failure
P')#
% ¡ a \" R EMOVE -F IRST( ) I'H§0G
¡ C # £
if G OAL -T EST[ T ¡  \" £
¤¤'¦
](S TATE[ ¡ a \"
')#
]) then return S OLUTION( ¡ a \"
')b# )
if S TATE[ ¡ a \"
')#
] is not in then a ¡  \"
2$@¥
add S TATE[ ] to )b#
¡ a \" 2$@¥
a ¡  \"
% ¡ C # £
`'I6§0G
I NSERT-A LL(E XPAND( , ), ) T ¡  \" £ ¡ a \"
¤¤2R ')# ¡ C # £
'I6§0G
Figure 3.25

7
Figure 4.13
£ \" D C ¡ # % © #¡££ 3
$2¢1§‰hRR§0§I¤¥
] © #¡££ 3
R¤2§I¥ if VALUE[neighbor] VALUE[current] then return S TATE[ j
a highest-valued successor of
© #¡££ 3
¢¤2§I¥ % £ \" D C ¡
V$24$@¤b#
loop do
]) T ¡  \" £
¤¤'¦ M AKE -N ODE(I NITIAL -S TATE[ % © #¡££ 3
R¢¤2§I¥
, a node £$24$@¤b#
\" D C ¡
local variables: , a node R§0§I¤¥
© #¡££ 3
inputs: , a problem ¤¤'¦
T ¡  \" £
function H ILL -C LIMBING( ) returns a state that is a local maximum
¤¤'¦
T ¡  \" £
Figure 4.6
if then return © 42¤2£
 3¡ 0I3 6$¤G v© I0¤2£
¡ £   uw  3  ¡
, [ ] RBFS( , ,
1$6$§¤© $i§6pH  ¤c@gƒ 0¤2 ¤¤2R
‘ ¡ d ©  # £ ¡   X © T G ˆ hf ©  ¡ T ¡  \" £ ) % 0§0 ‚ tI0§2£
©¡ © 3  ¡
the second-lowest -value among
£ \"¡¥¥ 3
§$¤0§§0¢0 ‚ Pe61b§¤¨t1
% ¡ d ©  # £ ¡ ©
if then return ©¡ , [
0§0 ‚ 2It6$¤G ]
¡ £ 3  6p6t g™0¤0 P‚
© T G ˜ —©¡ –
the lowest -value node in £ \"¡¥¥ 3
§$¤0§§0¢0 ‚ % ©¡
R0¤2
repeat
[s] ‘— ¡ a \" – X ‘ ˆ ” ’ ‘ ˆ ˆ ‡ ƒ
H¤')# P‚ §)s“•“)sbH‰1†„% ‚
for each in do £ \"¡¥¥ 3
§1¤0¤00¢0 
if is empty then return , y ¡2£43 6$¤G
  §$§0¤§2R2
£ \"¡¥¥ 3
E XPAND( , ) T¤¡¤'¦ ')#
 \"£ ¡ a \" c§$¤0§§0¢0
% £ \"¡¥¥ 3
if G OAL -T EST[ ]( ) then return¡ a \"
')b# ¡© ©
$Q0 ¤¤'¦ T ¡  \" £
‚
function RBFS( , , ) returns a solution, or failure and a new -cost limit
© T
Hp6  G ')b# ¤¤'¦
¡ a \" T ¡  \" £
RBFS( y
, M AKE -N ODE(I NITIAL -S TATE[
T ¡  \" £
§(§'¦ ]), ) ¤¤'¦
T ¡  \" £
function R ECURSIVE -B EST-F IRST-S EARCH( ) returns a solution, or failure
T ¡  \" £
§(§'¦
EXPLORATION 4
INFORMED SEARCH AND

8 Chapter 4. Informed Search and Exploration
function S IMULATED -A NNEALING( , T ¡  \" £
§(§'¦ ¡ 3 a ¡ D ¥
I$2¢§¤
) returns a solution state
inputs: ¤¤'¦
T ¡  \" £
, a problem
(I$¦¢¤¤
¡ 3 a ¡ D ¥
, a mapping from time to “temperature”
local variables: R§0§I¤¥
© #¡££ 3
, a node
, a node 10b#
© k¡
l
, a “temperature” controlling the probability of downward steps
M AKE -N ODE(I NITIAL -S TATE[ R¢¤2§I¥
% © #¡££ 3
¤¤'¦
T ¡  \" £ ])
for 1 to do y R©
%
[ ] © 431a2¢¤§g% l
¡ ¡ D¥
R§2£§43¤¥
© #¡ £ l
if = 0 then return
R#§2§43¤¥
© ¡££ a randomly selected successor of 10b#
% © k¡
©R#§¡2£§£43¤¥
VALUE[ ] – VALUE[ 10¡b#
© k ] % om n
if 0 then©$k2¡‰#hR©R§¡0§£I¤¥
% # £ 3 ˜ gm n
tsfrPip
s q else $2‰#hRR§0§I¤¥
© k¡
only with probability % © #¡££ 3
Figure 4.17
function G ENETIC -A LGORITHM( $!!$I04
# \" ©  3 \"
, F ITNESS -F N) returns an individual
inputs: $!1I§4
# \" ©  3 \"
, a set of individuals
F ITNESS -F N, a function that measures the fitness of an individual
repeat
empty set &$!$@I§4 ¤b#
% # \" ©  3 \" u ¡
$!!$I04
# \" ©  3 \"
loop for from 1 to S IZE( ) do
1\"!©$@4R§4
#  3 \"
R ANDOM -S ELECTION( vk
%
, F ITNESS -F N)
#$!$@I§4
\" ©  3 \"
R ANDOM -S ELECTION( w7
%
, F ITNESS -F N)
7 k
R EPRODUCE( , ) 6'§¥
% a D
a D
%v@6'¤¥
if (small random probability) then M UTATE( a D
@H¤¥ )
add to $!$@I§4 ¤‰#
# \" ©  3 \" u ¡ @6'¤¥
a D
$!$@I§4 ¤bgx$!!$I04
# \" ©  3 \" u ¡ # % # \" ©  3 \"
until some individual is fit enough, or enough time has elapsed
return the best individual in $!!$I04
# \" ©  3 \"
, according to F ITNESS -F N
7 k
function R EPRODUCE( , ) returns an individual
7 k
inputs: , , parent individuals
% &# L ENGTH( ) k
% `¥
random number from 1 to #
k
return A PPEND(S UBSTRING( , 1, ), S UBSTRING( , ¥ # {y¥ 7
z ’ , ))
Figure 4.21

9
function O NLINE -DFS-AGENT( ) returns an action | 
| 
inputs: , a percept that identifies the current state
static: © I0¤2£
 3¡
, a table, indexed by action and state, initially empty
a22£$@§2‰I3
¡ \" k¡ #
, a table that lists, for each visited state, the actions not yet tried
a2¡e9¤¦!1¤e2‰I3
¥ £© 9 ¥  #
, a table that lists, for each visited state, the backtracks not yet tried
 
, , the previous state and action, initially null
| 
if G OAL -T EST( ) then return §¨0
\"©
| 
if is a new state then % |  22$@00‰I3
a ¡ £ \" k ¡ #
[ ] ACTIONS( ) | 
if is not null then do 
[ , ] | h%   tI0§2£
 © 3  ¡
|  2¡e¤¦!©19¤¥e2‰I3
a 9¥ £
add to the front of  # [ ] 
if [ ] is empty then |  22$R§0bI3
a ¡ £ \" k ¡ #
§\"¨©2
if |  2e¤¦!1¤e2‰I3
a¡ 9 ¥ £© 9 ¥  #
[ ] is empty then return
else 3
|   © I0¤¡2£
an action such that  [ , ] = P OP( % w |  2e¤¦!1¤e2‰I3
a¡ 9 ¥ £© 9 ¥  #
[ ])
else P OP( [ ]) ¡ \" k¡ #
|  a22£$@§2‰I3 v
%
| h`
 %
return 
Figure 4.25
function LRTA*-AGENT( ) returns an action | 
| 
inputs: , a percept that identifies the current state
static: © I0¤2£
 3¡
, a table, indexed by action and state, initially empty
}
, a table of cost estimates indexed by state, initially empty
 
, , the previous state and action, initially null
| 
if G OAL -T EST( ) then return §¨0
\"©
|  } } |  ~% | 
”
if is a new state (not in ) then [ ] ( )
unless is null 
| h%   tI0§2£
[ , ] © 3¡
h
[ ] foƒ %  }
LRTA*-C OST( , ,  3¡
  © 42¤2£  
[ , ],
}
)
ƒs
‚ ¦
€ ACTIONS
 v
% |  }   © I0¤2£  
|  3¡ |
an action in ACTIONS( ) that minimizes LRTA*-C OST( , , [ , ], )
| h`
 %
return 
} |   
function LRTA*-C OST( , , , ) returns a cost estimate
|  iQc”
‘ „ˆ
if is undefined then return
else return — | „ – ˆg‘ | 02eQˆ
‡ ’ „X †X „
Figure 4.29

CONSTRAINT
5 SATISFACTION
PROBLEMS
¥

function BACKTRACKING -S EARCH( ) returns a solution, or failure
return R ECURSIVE -BACKTRACKING( , ) ¥ W‰
 Š
function R ECURSIVE -BACKTRACKING( , Q¥ ©R§¡gTRe@02$
# # C 
) returns a solution, or failure
if is complete then return R¤g¢e@001
© #¡ T # C 
©R§¡gRe@02$
# T # C 
S ELECT-U NASSIGNED -VARIABLE(VARIABLES[ ],  © #¡ T # C 
¥ R¤g¢e@001 Q¤¥
, ) % £ 
V$$d
for each in O RDER -D OMAIN -VALUES( , , ) do  © #¡ T # C  £ 
¥ R§gR$20$ $1d
3 
¡Rt1$d
if is consistent with R#§gR$20$
© ¡ T # C  Q¥

according to C ONSTRAINTS[ ] then ¡ 3 
Rt1$d
add = to #¡ T C 
©¢¤g¢#e@00$ Š R3 $$d £$1d ‰
¡ 
C 
Q¤¥ ©¢#¤¡gT¢#e@00$
R ECURSIVE -BACKTRACKING( , ) RtI0§2£
% © 3  ¡
if © I30¤2£
 ¡
then return 2£It6$¤G uvtI0§0£
¡ 3 w © 3  ¡
remove = T # C 
©¢#¤¡gR$@00$
from  £
Š ¡R3 $$d $$d ‰
return 2It6$¤G
¡ £ 3 
Figure 5.4
10

11
function AC-3( Q¥

) returns the CSP, possibly with reduced domains
inputs: 
Q¥
, a binary CSP with variables X    Ž
Š‘o‹ X g‹ X i‰
Œ ‹
local variables: R¤R¤S
¡ 3¡ 3
, a queue of arcs, initially all the arcs in Q¥

while §RS
¡ 3¡ 3 is not empty do
% “
R‘ p‹ X o‹ ˆ
’ R EMOVE -F IRST( ) ¡ 3¡ 3
R¤RS
if R EMOVE -I NCONSISTENT-VALUES( “ ‹ X’ ‹ ) then
for each in N EIGHBORS[ ] do
” †‹ ’ o‹
add ( ’ ‹ X ™‹
) to” R¤RS
¡ 3¡ 3
function R EMOVE -I NCONSISTENT-VALUES( “ ‹ X’ ‹ ) returns true iff we remove a value
 $¤•v2$$o¤2£
¡  G % a¡ d \" T¡
k
for each in D OMAIN[ ] do ’ ‹
if no value in D OMAIN[ ] allows ( , ) to satisfy the constraint between
– “ ‹ 7 k ’ ‹ and “ ‹
k
then delete from D OMAIN[ ]; ’ ‹ ¡R3§£©h%va2¡$d$\"o¤2£
T¡
return 2$$o¤2£
a¡ d \" T¡
Figure 5.9
function M IN -C ONFLICTS( , ¨0 ioT ¥
 ¡© k  
) returns a solution or failure
inputs: Q¥

, a constraint satisfaction problem
¨0 ioT
 ¡© k 
, the number of steps allowed before giving up
R¢¤2§I¥
% © #¡££ 3
an initial complete assignment for ¥

for = 1 to ¨0 ioT
 ¡© k 
do
¥
if is a solution for © #¡££ 3
¢¤2§I¥
then return © #¡££ 3
R§0§I¤¥
V$1d
% £ 
a randomly chosen, conflicted variable from VARIABLES[ ]¥

£ $1d d
the value for `3 $1d
% ¡  
that minimizes C ONFLICTS( , , 3, £ 
Q¤¥ ©R#¤¡2£§£I¥ d 1$d )
set R§¡2£§£43¤¥
© # in Rt1$d w 1$d
¡ 3  £ 
return 2It6$¤G
¡ £ 3 
Figure 5.11

6 ADVERSARIAL SEARCH
function M INIMAX -D ECISION( 1¨0
¡© ©
) returns $!¤—$
# \" ©¥  #
inputs: ¨$¨0
¡© ©
, current state in game
% wd M AX -VALUE(state)
return the # \" ©¥
$!¤
in S UCCESSORS( $Q0
¡© © ) with value d
function M AX -VALUE( 1¨0
¡© © ) returns Rt1$˜!6t6Ip
¡ 3  d 7©  © 3
if T ERMINAL -T EST( 1¨0
¡© © ¡© ©
¨$¨2
) then return U TILITY( )
y gwd
™ %
 X
R
for in S UCCESSORS( ) do¡© ©
$Q0
% wd M AX(v, M IN -VALUE( )) 
return d
function M IN -VALUE( $Q0
¡© © ) returns Rt$1˜!H 6Ix
¡ 3  d 7©  © 3
if T ERMINAL -T EST( ¡© ©
1¨0 ) then return U TILITY( ¡© ©
¨$¨2 )
y wd %
 X
R
for in S UCCESSORS( ) do¡© ©
$Q0
% wd M IN(v, M AX -VALUE( )) 
return d
Figure 6.4
12

13
function A LPHA -B ETA -S EARCH( ) returns an action ¨$¨2
¡© ©
inputs: ¨$¨0
¡© ©
, current state in game
% wd M AX -VALUE( , , ) ¡© ©
y ™ ¨$¨0 y ’
return the # \" ©¥
$!¤
in S UCCESSORS( $Q0
¡© © ) with value d
function M AX -VALUE( › š 1¨0
¡© ©
, , ) returns R3 $$p6t6!4x
¡   d 7©  © 3
inputs: ¨$¨0
¡© ©
, current state in game
, the value of the best alternative for
š MAX
1¨0
¡© ©
along the path to
, the value of the best alternative for
› MIN along the path to ¨$¨2
¡© ©
if T ERMINAL -T EST( ¡© ©
1¨0 ) then return U TILITY( ¡© ©
¨$¨2 )
y gwd
™ %
for  X
R in S UCCESSORS( ) do ¡© ©
$Q0
wd
% M AX(v, M IN -VALUE( , , )) › š 
›Ÿžœ if then return d
% šM AX( , v) š
d
return
function M IN -VALUE( › š $Q0
¡© ©
, , ) returns 3 $1˜6 6Ip
¡   d 7©  © 3
inputs: ¨$¨0
¡© ©
, current state in game
, the value of the best alternative for
š MAX
1¨0
¡© ©
along the path to
, the value of the best alternative for
› MIN along the path to ¨$¨2
¡© ©
if T ERMINAL -T EST( ¡© ©
1¨0 ) then return U TILITY( ¡© ©
¨$¨2 )
y ~wd
’ %
for  X
R in S UCCESSORS( ) do ¡© ©
$Q0
wd
%
M IN(v, M AX -VALUE( , , )) › š 
š jžœif then return d
› % ›
M IN( , v)
d
return
Figure 6.9

7 LOGICAL AGENTS
function KB-AGENT( §¦§¢
© ¡¥£¡
) returns an $!!
# \" © ¥
static: E, a knowledge base
¢¡
©
, a counter, initially 0, indicating time
T ELL( E ¢¡ , M AKE -P ERCEPT-S ENTENCE( , )) © ¨§¦¤¢
© ¡¥£¡
&$!!
% # \" © ¥A SK( E , M AKE -ACTION -Q UERY( ))
¢¡ ©
T ELL( E¢¡ , M AKE -ACTION -S ENTENCE( , )) © 1!
# \" © ¥
+1 h©
© %
#$!©¤¥
\"
return
Figure 7.2
function TT-E NTAILS ?( , ) returns
E ¢¡ or š ¡ 3£
R§© ¡  
t1G
inputs: , the knowledge base, a sentence in propositional logic
E ¢¡
š , the query, a sentence in propositional logic
 \" T 7
%P $2g10a list of the proposition symbols in E ¢¡ and š
return TT-C HECK -A LL( , , , )E x¡  \" T 7
— –  $2g10 š
function TT-C HECK -A LL( , , , E ¢¡ ¤')oT  $2g10 š
¡ a \"  \" T 7
) returns ¡ 3£
R§© or  $¤G
¡  
if E MPTY ?(  $2g10
 \" T 7
) then
if PL-T RUE ?( E ,
x¡ ¡ a \"
¤)oT
) then return PL-T RUE ?( , ¡ a \"
§')oT š )
else return R§©
¡ 3£
else do
% £ F IRST(  $2g10
);  \" T 7
R EST( ) % ©¡
R0§2£  \" T 7
 $2g10
return TT-C HECK -A LL( , , , E XTEND( 0¤2£ š ¢¡
©¡ E ¤')o§eR§§X £
¡ a \" TX ¡ 3£© ) and
TT-C HECK -A LL( , , ©0¤2£ š ¢¡
, E XTEND(¡ E ¤¡')o§¤t$¤HX £
a \" T X ¡   G )
Figure 7.12
14

15
function PL-R ESOLUTION( , ) returns E ¢¡or š ¡ 3£
R§© ¡  
 $¤G
inputs: , the knowledge base, a sentence in propositional logic
E ¢¡
š , the query, a sentence in propositional logic
%f¤I$@¥
 ¡  3  the set of clauses in the CNF representation of W€•¢¡
š ¥ ¤ E
ŠW‰ ¦§‰# % u¡
loop do
for each , in  ¡  3 
§I$@¥do
“ § ’ §
f!R¤$t1¤¤2£
% © # ¡ d \"  ¡
PL-R ESOLVE( , ) “ § ’ §
if R§ed $¤§0£
© #¡  \"¡
contains the empty clause then return R§©
¡ 3£
©¢#¤¡ed $¤§¡2£©¨—§‰h¦§‰#
\" u¡ # % u¡
¡ $¤G
if §I$@•«§‰#
 ¡  3  ¥ ª u ¡
then return
§‰#™c§I$@¬`¤¤41¤¥
u ¡ ¨  ¡  3  ¥ %  ¡  3 
Figure 7.15
function PL-FC-E NTAILS ?( , ) returns E ¢¡ or S R§!©
¡ 3£  $¤G
¡  
inputs: E, the knowledge base, a set of propositional Horn clauses
¢¡
S
, the query, a proposition symbol
local variables: RI$2¥
© # 3 \"
, a table, indexed by clause, initially the number of premises
22§§)6
a¡££¡ G #
, a table, indexed by symbol, each entry initially t1G
¡  
¤I
 a #¡ C
, a list of symbols, initially the symbols known to be true in E ¢¡
while eb¤')
 a #¡ C
is not empty do
P OP(b§I)
 a #¡ C ) % ­
unless 22§¤H
a¡££¡ G #
[ ] do
¡R3§£©g% 22§¤)6
[ ]a¡££¡ G #
for each Horn clause in whose premise ¥ appears do
decrement ¥ RI$0¥
[ ] © # 3 \"
if ¥ ¢410¥
© # 3 \"
[ ] = 0 then do
¥
if H EAD[ ] = then return S R§!©
¡ 3£
P USH(H EAD[ ], ) ¥ b§I)
 a #¡ C
return  $¤G
¡  
Figure 7.18

16 Chapter 7. Logical Agents
function DPLL-S ATISFIABLE ?( ) returns  ¡ 3£
R§!© or  $¤G
¡  

inputs: , a sentence in propositional logic
f¤I$@¥
%  ¡  3 
the set of clauses in the CNF representation of 
P $2g10
%  \" T 7
a list of the proposition symbols in 
return DPLL( , , ) — – t12g$2 §I$¤¥
 \"  T 7  ¡  3 
function DPLL( §')oT  $2g10 §I$@¥
¡ a \"  \"  T 7  ¡  3 
, , ) returns ¡ 3£
§!© or ¡  
¤t$¤G
if every clause in is true in  ¡  3 
¤I$@¥
then return ¡ 3£
R§© ¡ a \"
§')oT
if some clause in is false in  ¡  3 
§I$¤¥
then return ¡  
¤t$¤G ¡ a \"
§')oT
® PRt1$d
% ¡ 3  ¤')oT ¤¤41¤¥ t$¦g$0
¡ a \"  ¡  3   \"  T 7
, F IND -P URE -S YMBOL( , , )
® ® t12g$2 ¤I$@¥
 \"  T 7  ¡  3  ¤)o§Rt$1§X £
 ¡ a \" T X ¡ 3  d
if is non-null then return DPLL( , – , E XTEND( )
® PRt1$d
% ¡ 3  §')oT §I$@¥
¡ a \"  ¡  3 
, F IND -U NIT-C LAUSE( , )
® ® t12g$2 ¤I$@¥
 \"  T 7  ¡  3  ¤¡a)\"oT§X¡Rt$1d§X £
3
if is non-null then return DPLL( , – , E XTEND( )
% ® F IRST( ); % ©¡
R0¤2£  $2g10
R EST(  \" T 7
)  \"  T 7
t$¦g$0
return DPLL( , 0§2£ §I$¤¥
©  ¡  ¡  3 
, E XTEND( , , )) or §')oT R§© ®
¡ a \" ¡ 3£
DPLL( , 0§2£ §I$¤¥
©  ¡  ¡  3 
, E XTEND( , , )) ¤')oT  $¤G ®
¡ a \" ¡  
Figure 7.21
function WALK SAT( , ,  k   ¡  3 
 y¯ ioT §I$@¥
) returns a satisfying model or ¡ £ 3 
2It6$¤G
inputs: ¤I$@¥
 ¡  3 
, a set of clauses in propositional logic
, the probability of choosing to do a “random walk” move, typically around 0.5
  y¯ ioT
k 
, number of flips allowed before giving up
a random assignment of % ¡ a \"
e¤')oT
/ to the symbols in ¡   ¡ 3£
 $¤G R§!©  ¡  3 
¤I$@¥
for to  k 
R y¯ ioT
do z w
§¡I3$¤¥
if  satisfies ¡ a \"
§')oT
then return ¡ a \"
¤)oT
% ¡  3 
P¤41¤¥
a randomly selected clause from that is false in  ¡  3 
¤I$@¥ ¤')oT
¡ a \"
with probability flip the value in of a randomly selected symbol from ¡ a \"
§')oT I$@¥
¡  3 
else flip whichever symbol in I$¤¥
¡  3 
maximizes the number of satisfied clauses
return 2It6$¤G
¡ £ 3 
Figure 7.23

17
function PL-W UMPUS -AGENT( 02§¢
© ¡¥£¡
) returns an $!¤
# \" ©¥
inputs: © ¡¥£¡
R§¦¤¢
, a list, [ , £¡©©  ¡ °¡¡£ D¥ #¡©
¤¨6 eC i020§ ¤b§0
, ]
static: , a knowledge base, initially containing the “physics” of the wumpus world
E ¢¡
, , \" £
#$!©$Q©R#§¡Q§1\" 7 k
, the agent’s position (initially 1,1) and orientation (initially © D C
$@§£ )
2¨606id
a¡© 
, an array indicating which squares have been visited, initially  $¤G
¡  
$!¤
# \" ©¥
, the agent’s most recent action, initially null
$@
# 
, an action sequence, initially empty
update , , a¡©  # \" © © #¡ £
2¨606id $!1¨R§§$\" 7 k
, based on # \" © ¥
$!e
if then T ELL(
´ $c± ¢¡
³ ² E, ) else T ELL( , )¤¤2
D¥ #¡© ´ $²“—¥ ¢¡
³ ± E
if then T ELL(
´ $xµ ¢¡
³ ² , E ) else T ELL( , ) i002§
¡ °¡¡£ ´ ³$¢µ—¥ ¢¡
² E
if then '¦$~­$!¤
 £ C % # \" ©¥ ¤!6t$C
£ ¡©© 
%&#1\"!©¥
else if is nonempty then P OP( # 
$R
) # 
$@
else if for some fringe square [ , ], A SK(
¶ , ) is or ‘ ³“ ’ —„¤ ³“ ’ £ ¥ ˆ ¢¡
· ¥ ¡ 3£
R§!©
E
for some fringe square [ , ], A SK(
¶ , ) is ¡t1G
then do ‘ ³“ ’ ·  ³“ ’ £ ˆ ¢¡
¸ E
% # 
&$R
A*-G RAPH -S EARCH(ROUTE -P ROBLEM([ , ], #$!$Q©R§Q§1\" 7 k
\" ©  #¡ £
, [ , ], ¶ a¡© 
¦606id ))
P OP( ) $@
#  ­$!e
% # \" © ¥
else a randomly chosen move &$!!
% # \" © ¥
return $!¤
# \" ©¥
Figure 7.26

8 FIRST-ORDER LOGIC
18

9 INFERENCE IN
FIRST-ORDER LOGIC
¹ 7 k
function U NIFY( , , ) returns a substitution to make and identical k 7
k
inputs: , a variable, constant, list, or compound
7
, a variable, constant, list, or compound
, the substitution built up so far (optional, defaults to empty)
¹
if = failure then return failure
¹
k
else if = then return 7 ¹
k
else if VARIABLE ?( ) then return U NIFY-VAR( , , ) ¹ 7 k
7
else if VARIABLE ?( ) then return U NIFY-VAR( , , ) ¹ k 7
else if C OMPOUND ?( ) and C OMPOUND ?( ) then k 7
return U NIFY(A RGS[ ], A RGS[ ], U NIFY(O P[ ], O P[ ], ))k 7 k 7 ¹
k
else if L IST ?( ) and L IST ?( ) then 7
k
return U NIFY(R EST[ ], R EST[ ], U NIFY(F IRST[ ], F IRST[ ], )) 7 k 7 ¹
else return failure
function U NIFY-VAR( , , ) returns a substitution ¹ k $1d
£ 
inputs: $$d
£ 
, a variable
k
, any expression
, the substitution built up so far
¹
if ¹ U»¦Š   d º £ 
$$I$$1d ‰
then return U NIFY( , , ) ¹ k $$d
 
¹U»¦Š $14º )‰
else if  d ¼ then return U NIFY( , , ) ¹   
$1d £$$d
else if O CCUR -C HECK ?( k $$d
£ 
, ) then return failure
else return add / to ¹ Š k $$d ‰
£ 
Figure 9.2
19

20 Chapter 9. Inference in First-Order Logic
function FOL-FC-A SK( , ) returns a substitution or E x¡ š ¡  
t1G
inputs: , the knowledge base, a set of first-order definite clauses
E ¢¡
, the query, an atomic sentence
š
local variables: ¤b#
u¡
, the new sentences inferred on each iteration
repeat until §‰#
u¡ is empty
Š W‰ —¤b#
% u¡
for each sentence Ex¡ £ in do
%R$S ‘ ½•‘ ¤  `Œ §ˆ
  ¤
S TANDARDIZE -A PART( ) £
¤  ¤ Œ ¹
for each such that S UBST( ,¹ ) = S UBST( , ‘ ¤ |Œ ¹  
| ‘ ¤ ) )
Ex¡ | ‘ sXX |Œ
 for some in
S ¹
S UBST( , ) % | S
| S
if is not a renaming of some sentence already in or E ¢¡ u¡
¤b# then do
add to §‰# | S
u¡
š | S
U NIFY( , ) % ¾
¾ 6$¤G
  ¾
if is not then return
add to ¢¡
E ¤‰#
u¡
return  $¤G
¡  
Figure 9.5
function FOL-BC-A SK( , , ) returns a set of substitutions
E ¢¡   \"
¹ t$)C
inputs: , a knowledge base
E ¢¡
 $)'C
  \"
, a list of conjuncts forming a query ( already applied) ¹
, the current substitution, initially the empty substitution
¹ Š W‰
local variables: §¤y0R1
£¡ u #
, a set of substitutions, initially empty
 $)C
if  \" is empty then return '1‰
Š ¹
% | S
S UBST( , F IRST( )) ¹  $)C
  \"
for each sentence in where S TANDARDIZE -A PART( ) =
E
¢¡ £ £ €Œ §ˆ
¤ ‘ Àž‘ ¤ )
¿ ½  
% | ¹
and U NIFY( , ) succeeds | S S
%   \" u ¡
ft1)'C ¤b# R EST( ) Á X   
f‘ sX Œ – —  $)'C
  \"
% £¡ u #
f§¤y0R1
FOL-BC-A SK( , , C OMPOSE( , )) E ¢¡  $)\"C §‰#u¡ ¹ | ¹ £¡ u # 
§§y0R$¨
return §¤y0¢$
£¡ u #
Figure 9.9
procedure A PPEND( # \" ©  3 # © # \" ° k
$!$¢R6!R$0¥ i 7 )
, , , )
6$¦©
%  £ G LOBAL -T RAIL -P OINTER()
if k ) —–
= and U NIFY( , ) then C ALL( ) ° i 7 # \" ©  3 # © # \"
$!1R¢6R10¥
R ESET-T RAIL( ) 6$¦©
 £
% v N EW-VARIABLE(); N EW-VARIABLE(); N EW-VARIABLE() % Ãk % y°
k i  k
if U NIFY( , [ — ]) and U NIFY( , [ — ]) then A PPEND( , , , ° i  ° $!!$¢R6¢$0¥ ° 7 k
# \" ©  3 # © # \" )
Figure 9.12

21
procedure OTTER( , ) ¤'§43 $¤
¡     \"
inputs: $¤
 \"
, a set of support—clauses defining the problem (a global variable)
¤'§43
¡   
, background knowledge potentially relevant to the problem
repeat
%
clause the lightest member of $§
 \"
move from ¡  3 
I$@¥
to ¡   
¤¤I3  \"
$¤
P ROCESS(I NFER( , ), $\"¤ ¤'§43 I$@¥
) ¡    ¡  3 
until $¤
 \" —–
= or a refutation has been found
function I NFER( (§'¤I3 I$@¥
¡    ¡  3 
, ) returns clauses
resolve I$@¥
¡  3 
with each member of ¤'§43
¡   
return the resulting clauses after applying F ILTER
procedure P ROCESS( $¤ ¤I$@¥
 \"  ¡  3 
, )
¡  3 
I$@¥
for each in do  ¡  3 
¤I$@¥
P¤41¤¥
% ¡  3 
S IMPLIFY( I$@¥
¡  3 
)
merge identical literals
discard clause if it is a tautology
[%  \"
`$§ — I$@¥
¡  3 
]  \"
$§
I$@¥
¡  3 
if has no literals then a refutation has been found
¡I$@¥
if  3 has one literal then look for unit refutation
Figure 9.19

10 KNOWLEDGE
REPRESENTATION
22

23
Figure 11.3
‘ ‘ \"© X ̈ A
¦'Qy“@¤© ɤ †$¦0b“@¤© ¦¥ E FFECT:
‘ T \" £ G X ̈ A
‘ \"©ˆ©£ \" £ A ‘ T \" £ Gˆ © £ \" £ A
¨s¤§$4¨6 Ǥ †$¦§@¤§$4¨6 Ǥ @'b$@ ® ¤ ™120c@¤© A ‘ ̈ ¡ #  ‘ T \" £ G X ̈ P RECOND :
§'Qy€$¦0b“@b7  Å P$!¤¥ A
X ‘ \"© X T \" £ G X ̈ ˆ # \" ©
‘‘ Ì ˆ ¥
¦X P# Ä „¤ 4PX ¤© A E FFECT:
‘ † ˆ
‘ †ˆ©£ \" £ A ‘ ̈ ¡ #  ˆ \" C£ Ê
4¤§$4¨6 Ǥ b@‰$@ ® ¤ ‘ ¢'$ ˆ¤ I`“@¤© Ǥ bX P# Ä ‘ † X ̈ A ‘ Ì ˆP RECOND :
§4P“X “)@Rf$P$!¤¥ A
X ‘ † X Ì ˆ a  \" # Í ˆ # \" ©
‘‘ Ì ˆ
¦X P# Ä ¤ 4PX ¤© ¦¥ E FFECT:
‘ † ˆ A
£ \" £ A ‘ ̈ ¡ #  ˆ \" C£ Ê
‘4†ˆ¤©§$4¨6 Ǥ @'b$@ ® ¤ ‘ ¢'$ ˆ¤ 4`“@¤© Ǥ IPX ¤© A ‘ † X ̈ A ‘ † ˆ P RECOND :
X‘ † X Ì ˆ a  ˆ # \" ©
§4`“bX “e)\" F P$!¤¥ A
‘ Æ Å Ž ˆ A ¡ È Œ ˆ ˆ 
¦‘ vI8 X v§ ¤© ɤ ‘ PÅ yX v§ ¤© A 1)\" Ë
‘ Æ Å ˆ ©£ \" £ A
¦‘ w48 §14¨H ɤ ‘ PÅ ¦¤§$4¨6 ¦¤ ¡ Ȉ©£ \" £ A
ˆ ¡ #  ˆ ¡ #  Ž
‘ Ž £ ‰$@ ® ¤ ‘ Œ £ ‰1 ® ¤ ‘ v§ ¢Q1 Ÿ¤ ‘ v§ R$ •¤ˆ \" C£ Ê Œ ˆ \" C£ Ê
¡ È ˆ A Æ Å ˆ A ¡ È ˆ A
‘ PÅ yX Ž £ ¤© ɤ ‘ w48 X Œ £ ¤© Ǥ ‘ PÅ yX Ž § ¤© Ǥ ‘ v48 X Œ § ¤© A 6¢# Ä Æ Å ˆ ˆ ©
PLANNING
11

Figure 11.7
‘ Æ ˆ £ ¡ Ê
¦‘ ¼ XiÑQˆ`# •¥„¤ ‘ ¼ I$2 Ÿ¤ 1¤ l iQ`# Æ ‘ ¡  X ш E FFECT:
, ‘ ¼ u{Qˆ ¤ ¤Qˆb¤)@ Eɤ ¤ÑQI$2 Ÿ¤ ‘ ¼ iQP# Æ
w Ñ ‘Ñ 9¥ \" ‘ ˆ £ ¡ Ê X ш P RECOND :
X X ш ¡ 
§‘ ¼ iQ(§' l \" l e1ШP$!¤¥ A ¡ d \" ψ # \" ©
¥
‘ ˆ £ ¡
¦‘ – I$2 ʕɤ ‘ ¼ iÑQP# •¥„¤ ‘ ¼ ˆ412( ˆ¤ ‘ – iQ`# Æ
X ˆ Æ £ ¡ Ê X ш E FFECT:
,‘ – u w ¼ ˆ ¤ ‘ – &Qˆ ¤ ‘ ¼ {Qˆ
uw Ñ uw Ñ
¤ ‘¤Qb9§) Eɤ ‘ – 4$¦( ˆ¤ ‘¤QˆI$2¡ Ÿ¤ ‘ ¼ iQP# Æ
ш ¥ \" ˆ £ ¡ Ê Ñ £  Ê X ш P RECOND :
X §‘ – X ¼ iQe1ШP$!¤¥ A
X ш ¡ d \" ψ # \" ©
‘
¦‘ § X µ `# ˆ¤ ‘ µ &P# Æ 1)\"
ˆ Æ X Έ ˆ 
‘ ˆ £ ¡ Ê
¦‘ § 4$¦( ˆ¤ ‘ µ I$2 Ÿ¤ x4$¦( ˆ¤ Ë
ˆ £ ¡ Ê ‘ Έ £ ¡ Ê
‘ § b¤)@ ɤ ‘
ˆ 9¥ \" E µ ˆ 9¥ \" E
b¤)@ ɤ xb¤)@ ɤ ‘ Έ 9¥ \" E
‘ ¡  ˆ Æ ‘ ¡ 
i(§' l X § P# ˆ¤ i¤' l X ˆ Æ
µ P# ˆ¤ 1¤ &`# Æ 6¢# ‘ ¡  l X Έ ˆ © Ä
Figure 11.5
‘ ‘ ¡ X © 
¦i(1k A §1 Å ˆ A ¥
¤© ¦„¤ R412£ §$ Å ¤© ¦Y¤
‘ a # 3 \" X©  ˆ A ¥
‘ 9 # 3 X ¡£  ˆ A ¥ ‘ ¡ X ¡ £ 
444§£ l 2$4 8 ¤© ¦„¤ i$k A 2$4 8 ¤© ¦„¤ R412£ Ë 2$4 8 ¤© ¦¥
ˆ A ¥ ‘ a # 3 \" Ë X ¡£  ˆ A E FFECT:
P RECOND :
,
© D C #£¡
'1@R§¤$d Æ $$2¡ F P$!¤¥ A
¡ d  ˆ # \" ©
‘ ¡ X ¡ £  ˆ A ‘ a # 3 \" X ¡£  ˆ A
¦‘)(1k A e2$4 8 ¤© ɤ R412£ Ë 2$4 8 ¤© ¦¥ E FFECT:
‘  X©  ˆ A ¥ ‘ a # 3 \" X ¡£  ˆ
)¡(1k A §$@ Å ¤© ¦É¤ RbI$¦£ Ë 2$4 8 ¤© A P RECOND :
‘ ¡, X ¡£  ˆ © ˆ # \" ©
i$k A 2$4 8 P# Æ I3 ® P$!¤¥ A
‘‘ # \" X©  ˆ A ‘ ¡ X ©  ˆ A
¦¢abI3$¦£ Ë §$@ Å ¤© Ǥ )(1k A §$@ Å ¤© ¦¥ E FFECT:
‘ ¡ X ©  ˆ
i$k A ¤1 Å ¤© A P RECOND :
,
‘ ¡ X ©  ˆ ¡ d \" T ˆ # \" ©
1$k A §$ Å ee$o§¡ B P$!¤¥ A
a \" ˆ A ‘ 9 # 3 X ¡£  ˆ A
‘¦‘¢b#I3$¦£ Ë X¡2£$4 8 ¤© Ǥ ¢4I§£ l e2$4 8 ¤© ¦¥ E FFECT:
‘ 9 # 3 X ¡£  ˆ
4¢I§£ l 2$4 8 ¤© A P RECOND :
,
‘ 9 # 3 X ¡£  ˆ ¡ d \" T ˆ # \" ©
444§£ l 2$4 8 ee$o§¡ B P$!¤¥ A
‘ ‘ ¡ X ¡ £  ˆ ˆ 
¦)(1k A e014 8 ¤© A s1)\" Ë
‘‘ 9 # 3 X ¡£  ˆ A ‘ ¡ X ©  ˆ ˆ ©
¦4¢I§£ l 2$4 8 ¤© ɤ i$k A §$@ Å ¤© A 6¢# Ä
Planning 11. Chapter 24

Figure 11.19
) E XPAND -G RAPH(
T ¡  \" £ D  £
¤¤'¦ R02$C , ÃR§¦$C
% D £
0I3 6$¤G
¡£   ) then return D £
R02$Celse if N O -S OLUTION -P OSSIBLE(
$!It1¤
# \" © 3 \" if then return w \" ©  \"
¡0£I3 6$¤G uÐ#$!!43 $¤

)) ¢§¦eC
D £ , , L ENGTH(
 $)C R§¦eC
  \" D £ E XTRACT-S OLUTION( &$!I3 $¤
% # \" ©  \"
then do D £
R02$C all non-mutex in last level of   \"
t1)'C if
loop do
G OALS[ ] ¤¤2R
T ¡  \" £ %   \"
` $)'C
) T ¡  \" £
§(§'¦ I NITIAL -P LANNING -G RAPH( % D £
vR§¦eC
) returns solution or failure ¤¤'¦
T ¡  \" £ function G RAPHPLAN(
Figure 11.16
‘ ¡ 9 ˆ ¡ d
ie1 Ê e$ } E FFECT:
¡ d
‘)¡91 Ê ˆ$$ } ¥ P RECOND :
)1 Ê i E P$!¤¥ A
‘ ¡ 9 ˆ ¡ 9 ˆ # \" ©
) ‘ 9 ˆ ¡ d
‘1¡e91 Ê ˆP#§¡©$ n ¤ i¡e1 Ê e$ } ¥ E FFECT:
‘ ¡
)91 Ê ˆ$$ } ¡ d P RECOND :
¡ 9 ˆ © ˆ # \" ©
‘ii Ê $ n P$!¤¥ A
‘‘ ¡ 9 ˆ #¡©
¦ii Ê P¤¨$ n ¤ ‘ ¡ 9 ˆ ¡ d ˆ 
)1 Ê $$ } 1)\" Ë
‘‘ ¡ 9 ˆ ¡ d ˆ ©
¦i1 Ê ee$ } 6¢# Ä
Figure 11.11
‘ ‘ ¡ X © 
¦i(1k A §$@ Å ˆ A ¥
¤© ¦„¤ R412£ §$ Å ¤© ¦Y¤
‘ a # 3 \" X©  ˆ A ¥
‘ 9 # 3 X ¡£  ˆ A ¥ ‘ ¡ X ¡ £ 
444§£ l 2$4 8 ¤© ¦„¤ i$k A 2$4 8 ¤© ¦É¤ R412£ Ë 2$4 8 ¤© ¦¥
ˆ A ¥ ‘ a # 3 \" Ë X ¡£  ˆ A E FFECT:
P RECOND :
,
'1@R§¤$d Æ $$2¡ F P$!¤¥ A
© D C #£¡ ¡ d  ˆ # \" ©
‘ ¡ X ¡ £  ˆ A ‘ a # 3 \" X ¡£  ˆ A
¦‘i$k A e2$4 8 ¤© Ǥ R412£ Ë 2$4 8 ¤© ¦¥ E FFECT:
‘  X©  ˆ A ¥ ‘ a # 3 \" X ¡£  ˆ
)¡(1k A §$@ Å ¤© ¦É¤ RbI$¦£ Ë 2$4 8 ¤© A P RECOND :
, ‘ ¡ X ¡£  ˆ © ˆ # \" ©
i$k A 2$4 8 P# Æ I3 ® P$!¤¥ A
‘‘ # \" X©  ˆ A ‘ ¡ X ©  ˆ A
¦¢abI3$¦£ Ë §$@ Å ¤© Ǥ )(1k A §$@ Å ¤© ¦¥ E FFECT:
‘ ¡ X ©  ˆ
i$k A ¤1 Å ¤© A P RECOND :
,‘ ¡ X ©  ˆ ¡ d \" T ˆ # \" ©
1$k A §$ Å ee$o§¡ B P$!¤¥ A
a \" ˆ A ‘ 9 # 3 X ¡£  ˆ A
‘¦‘¢b#I3$¦£ Ë X¡2£$4 8 ¤© Ǥ ¢4I§£ l e2$4 8 ¤© ¦¥ E FFECT:
‘ 9 # 3 X ¡£  ˆ
4¢I§£ l 2$4 8 ¤© A P RECOND :
, ‘ 9 # 3 X ¡£  ˆ ¡ d \" T ˆ # \" ©
444§£ l 2$4 8 ee$o§¡ B P$!¤¥ A
‘ ‘ ¡ X ¡ £  ˆ ˆ 
¦)(1k A e014 8 ¤© A s1)\" Ë
‘‘ 9 # 3 X ¡£  ˆ A ‘ ¡ X ©  ˆ ˆ ©
¦4¢I§£ l 2$4 8 ¤© ɤ i$k A §$@ Å ¤© A 6¢# Ä
25

26 Chapter 11. Planning
§(§'¦
T ¡  \" £ ¦P˜l
ÔÓ Ò
function SAT PLAN( , ) returns solution or failure
inputs: ¤¤'¦
T ¡  \" £
, a planning problem
¦P˜l
ÔÓ Ò
, an upper limit for plan length
¦P˜l
ÔÓ Ò l
for = 0 to do
ÃI60oT ¥
% C #  G # §(§'¦
T ¡  \" £ l
, T RANSLATE -T O -SAT( , )
R¢¤gR$@00$
% © #¡ T # C 
SAT-S OLVER( ) ¥
G #
if R§gR$20$
© #¡ T # C 
is not null then
return E XTRACT-S OLUTION( '6§oT R§gRe@02$
C #  © #¡ T # C 
, )
return 2It6$¤G
¡ £ 3 
Figure 11.22

27
Figure 12.2
E FFECT:
àà ë òÙ
§¢$0vÖ × Ø Ü ¢ ö á à û Ù ç
õ QI£¡ÂP$R`Ö õ ö
P RECOND :
Ý Ý Ý Û á à Ù Ö Ýð ç Û í à ûÙ ÖÕ ç Ö× æ Ö
éà1ûÙ¢!×'ÜbyÚâW1ûvæ§tiçbЄáP1¤fIRxä é à ûÙ ç ¨Ý Ö ÕÙ
1iØ ø “©‰svÖ õ QØ x÷
× ø
E FFECT:
à þÙ
§àb¨wÖ × Ø Ü ¢ ö á à û Ù Ö ¦ Ý ð ç ç Û
õ QI¤£¡ÂW1vçtibÐí
P RECOND :
à û Ý×Ý Ü Û
é1Ù¢¤Ý'bwÚ«Wàb¤e)ég¥RtðibÛÐí
á þé û ¤Ù Ý ç ç 1†¥¢tibÐí ¢x÷ vÖ
é à û é ¤ Ù Ýð ç ç Û ù ù Ù õ QØ x÷
× ø
E FFECT:
àà þ
§¨ÙvÖ õ ×QI£¡åW1v¤`'xä
Ø Ü ¢ ö á à ûÙ Ö Õç Ö× æ Ö
P RECOND :
é à Ù Ö Õ Ö× Ö ä
1ûv¤ç`'æxÿåW1û¢!'ÜbyÚâáPàb¤e)I`!Ixä
á à Ù Ý× Ý Ý Û þ é û é úÙ ç Ö× æ Ö
é à üé û é úÙ ç Ö× æ Ö ù ù Ù
ý¤eIR`'xä ¢x÷ vÖ õ QØ x÷
× ø
àà ã ÞÙ ç ö á à ß ÞÙ ç
§$­fÖ õ ÂW)xR`Ö õ s2IÜ &ô öÙð õ
à§à¢e$'¢)•s¢@ibÐîWRI1p$Us¢t1bïîá
ó ò é ã Þ é ã ñ Ù Ýð ç ç Û í á à ë ê é ß Þ é ß ñ Ù Ýð ç ç Û í
à ë ì é ã Þ é ã èÙ ç Ö× æ Ö ä á à ë ê é ß Þ é ß èÙ ç Ö× æ Ö ä
RI­p¨`!IÃÂWR')1p1g¨fIRxåá
à ã ÞÙ Ý× Ý Ý Ü Û Ú á à ß ÞÙ Ý× Ý Ý Ü Û Ú Ù Ø× Ö
$­4!'byâW)p4¤'bw1iQ“¤Õ
IN THE REAL WORLD 12
PLANNING AND ACTING

Figure 12.9
‘ D # 3
¦‘ Š ¦H¢6 Å C20 d4 ™ ± ‰26R6 Å 7 b©5 ™ \"±
8 7 5  c ! X D  # X W U ! 3
X 3
X ! ± EaG W 97\"™ ©5 V ± X  ± CTRQP0 ™ 5 IŽ ± X  ± G FC\"7 ™ AŒ ±
` 3 YW U G S ( B G\" H E D 3 CB @
! 5
X ± 90 £™ '§$¨© 8 X f± 2©\"™ '§1¨© 'e
8 7 3 64 © £  Œ 1 0 3 )& © £  8 ‰ ( L INKS
Š 
X i± $ Ž “±  !
$¡©§£$Q© 8 X‰D26R6# Å $ ± $ ± f± %§1¨© 'e  Œ $ ©£ 8‰ O RDERINGS
Š £ ¡ a ! $
§'@HI3 E 7$ ® #\"± f$!¥¢§!0R$\" † ± X # \" © 3 £© # Ê 
I¤6I3 E ¡26 “± 06pT§¤¡ ©
X £¡ a £ Ž X© £  Œ ±‰
`$e
S TEPS ˆ # 
} ® ¤¡ Ë P$@ ®
5
X ¡ 3 a
I$\" } @643 E '$4Ã$22¡ ˆ ¡  \" T \"¥
‘©¥  © #
2¤¦£¢$\" ʕ„¤ I3$\" } ¤ ¤‰1ÐÏ ¥ ¥ ¡ 7¡ # \" E FFECT:
X  ¡ 3
¤© 643 E I$\" } ¤ ¤b$\"ÐÏ 7¡ # P RECOND : X £ ¡ a 7 ˆ # \" ©
4§'@HI3 E $ ® P$!¤¥ A
‘© £
¦6pT§¤¡ ® ¥„¤ © HI3 E I$\" } E FFECT:
¡  3
X© ¥ £© # Ê
§¦!R$\" ˆ¤ Hp§§¡ ® © T£ P RECOND : X`#$!¤¥¢§©0¢$\"
\" © 3£  # Ê P$!¤¥ˆ # \" © A
E FFECT:
¦¤¦¢$\" Ê
‘©¥ £© # X ¡ a
I£§'6I3 E 06 ¡£ ˆ # \" © A
E FFECT: P RECOND :
} P$!¤¥ A
©
‘¦6pT§£¤¡ ® ‰b$ F
X a # X© T
0Hp§£§¡ ® ¤¡ © Ë P$!¤¥ ˆ # \" ©
‘ ¡  3
iI$\" } E FFECT: X a #
‰b1 F P RECOND : 3
X¡¤41\" } 
a@6I3 E P$!¤¥
ˆ # \" © A
‘ ¡ C  C© £ \"
)I)6§$ÐÏ ¤ ¤‰1ÐÏ 7¡ # \" E FFECT: X© a¡ a \"
¤6$22£ Ê ))\" Ë P RECOND : X # 
`$)\" F ¤¡
© ˆ # \" © A
¥
I¤‰1ÐÏ €¤ b$ F
‘ 7¡ # \" a # E FFECT: X 7¡ # \"
§‰$ïÏ P RECOND : “$
X a # F I3 7
Ë E P$!¤¥
ˆ # \" ©
P$!¤¥ A
Figure 12.5
R ESOURCE :
àà òÙ Ý
§04)¢ õ Ø ø “©“¤Õ
ç ¨Ý Ö
E FFECT:
é à ë òÙ ×Ø Ü ¢ ö á à ûÙ ç
¢$0vÖ õ QI£¡ÂP$R`Ö õ ö
P RECOND :
1vçt1bï„W1v¤`!Ixä
é à û Ù Ö ¦ Ýð ç ç Û í á à û Ù Ö Õ ç Ö× æ Ö
é à ûÙ ç ¨Ý Ö ÕÙ × ø
$iØ ø “©‰swÖ õ QØ x÷
R ESOURCE :
àà òÙ Ý ×Ø ÜØ ð çç Û
§0R‰Ö õ QIsf 1¤bÐí E FFECT:
é à þ Ù × Ø Ü ¢ ö á à û Ù Ö ¦ Ýð ç ç Û
b¨wÖ õ QI¤£¡ÂP1Ã§tibÐí
P RECOND :
1¢!'byâWe†¥¢@ibÐí
é à û Ù Ý× Ý Ý Ü Û Ú á à þ é û é ¤Ù Ýð ç ç Û
é à û é ¤ Ù Ýð ç ç Û ù ù Ù × ø
1†¥¢tibÐí ¢x÷ wÖ õ QØ x÷
R ESOURCE :
àà òÙ ÝØ Ý
§0RQ× õ `'xä
ç Ö× æ Ö E FFECT:
é à þÙ ×Ø Ü ¢ ö á à ûÙ Ö Õç Ö× æ Ö
¨vÖ õ QI£¡åW1v¤`'xä
P RECOND :
é à ûÙ Ö Õ ç Ö× æ Ö ä ÿ á à ûÙ Ý× Ý Ý Ü Û Ú á à þ é û é úÙ ç Ö× æ Ö
1v¤`'xåW14¤'bwâPb¤e)I`!Ixä
é üé û é úÙ ç Ö× æ Ö ù ù Ù × ø
à §e)IR`'xä ¢x÷ wÖ õ QØ x÷
àà ã ÞÙ ç ö á à ß ÞÙ ç
§$­R`Ö õ ÂPixR`Ö õ s2IÜ &ô öÙð õ
ÖÕ á à òÙ Ý ×Ø ÜØ ð ç ç Û í á à òÙ ÝØ Ý
à§à¢©Ù¢Ý)¢ õ Ø ø 瓨©Ý“¤åW0¢“Ö õ QIsf ibЄP0RQ)!× õ fIRxåá  ç Ö× æ Ö ä
à¢e$'¢)•s¢@ibÐîWRI1p$Us¢t1bïîá
ó ò é ã Þ é ã ñ Ù Ýð ç ç Û í á à ë ê é ß Þ é ß ñ Ù Ýð ç ç Û í
à ë ì é ã Þ é ã èÙ ç Ö× æ Ö ä á à ë ê é ß Þ é ß èÙ ç Ö× æ Ö ä
RI­p¨`!IÃÂWR')1p1g¨fIRxåá
à ã ÞÙ Ý× Ý Ý Ü Û Ú á à ß ÞÙ Ý× Ý Ý Ü Û Ú Ù Ø× Ö
e¢¢!'byâW)p4!'by11“Õ
Planning and Acting in the Real World 12. Chapter 28

Figure 12.18
return P OP( ) $@
# 
A PPEND(
#1\"!$¢3R6©¢$\"0¥ 6$4¨¡0£
©  # # £  , )
% #  % #  ¡ \" D
&1î&$@ (14Pu
„ the tail of $@ $4Pu
#  ¡ \" D starting at &$!$¢¢H!R$2¥
% # \" ©  3 # © # \"
E¢¡ P LANNER(
„ ©R§¡0£§£I¤¥
# 3 , , )
% £  ¡ uw ¡ £ 3  
V6$42£ x2It6$¤G
find state in such that ¡©  a a # 
§$!$$0¥ „
) candidates S ORT(
© #¡££ 3
¢¤2§I¥ , ordered by distance to
$@ $4Pu
#  ¡ \" D %
E ¢¡
if P RECONDITIONS( F IRST( )) not currently true in
# 
$R then
P LANNER(
x¡ $)'C ¢¤2§I¥
E   \" © #¡££ 3 , , ) %&#1 î%&#$@ ¡(14Pu
   \" D
if then — – w $R # 
S TATE -D ESCRIPTION( , )
© E ¢¡ R¢¤2§I¥
% © #¡££ 3
, M AKE -P ERCEPT-S ENTENCE(
© ¨§¦¤¢
© ¡¥£¡ , )) E
¢¡ T ELL(
, a goal 1)'C
  \"
, a plan, initially
—– #1 $4Pu
 ¡ \" D
, a plan, initially —– $@
# 
static: , a knowledge base (includes action descriptions) E ¢¡
function R EPLANNING -AGENT(
# \" © ¥
$!! ) returns an
§¦§¢
© ¡¥£¡
Figure 12.14
‘ else Œ e
cf‘ then Œ e if else Ž Ž
return if then Œ Œ
else if then
— # 
$@ # 
1 cf‘  
  #$@
  $@
#   –
¡ £ 3 
2It6$¤G if = then return
243 6$¤G
¡£   $@
# 
) H©$4 ¤¤'¦ ’ 
D T ¡  \" £ O R -S EARCH( , ,% ’ $R # 
for each in do
§ ¨$¨0
©¡ ¡© © ’
¡ £   G £ \" X #    # \" © a # \" ¥
2I3 H1P$v`$@¢1b$!6$$0˜ ) returns D© 
!14 , function A ND -S EARCH(
§(§'¦ § ¨$¨0
T ¡  \" £ © ¡ ¡© © ,
return 2It6$¤G
¡ £ 3 
if then return
# 
—b$@ Á #$!¤ –
\" ©¥ 243 6$¤G Ð$@
¡ £   uw #  
A ND -S EARCH( ,
—4!14 Á ¡©$Q0 – T¤¤'¦£ ¤¤ $Q0
D©  © ¡ \" ©¡ ¡© © , ) &$R
% # 
for each
¡© © ,
$Q0 ¤¡¤'\"¦
T  £ in S UCCESSORS[ ]( ) do
©¡ ¡© © # \" ©¥
¤¤ $Q0 $!¤
if is on then return 3 
¡2£It6$¤G !$4
D©  ¡© ©
¨$¨2
if G OAL -T EST[ ]( ) then return the empty plan
$Q0 §(§'¦
¡© © T ¡  \" £
2It6$¤W$vf$R¢$$!6$$0o function O R -S EARCH(
¡ £ 3  G £ \" X #    # \" © a # \" ¥ , ,
D© 
!$4 ) returns
§(§'¦ ¨$¨0
T ¡  \" £ ¡© ©
—– , )T ¡  \" £ ],
¤¤'¦ ¤¤'¦ O R -S EARCH(I NITIAL -S TATE[
T ¡  \" £
243 6$¤W$v`$@4$b1!6$b$2o
¡ £   G £ \" X #    # \" © a # \" ¥ ) returns T ¡  \" £
§(§'¦ function A ND -O R -G RAPH -S EARCH(
29

Figure 12.30
‘‘— Ñ X X © # ¡ C  ˆ A ¥
¦62WR† – 0R§I)2¤© ¦„¤ 6— – X –¼ §¢¤I2¤© A
‘ E FFECT:
X© #¡ C ˆ
X ‘—Ñ X X © # ¡ C ˆ
§H¦WR† – 0R§I)2¤© A P RECOND :
X‘ X© #¡ C ˆ ˆ # \" ©
§H— – X –¼ 0R¤')2R\" Ë P$!¤¥ A
‘‘ ˆ a ¡ # £ 3©
¦!6$ E ‰2‰§I!¤¡ B E FFECT:
‘ X £ ¡ #© £  ˆ A ¥ ‘ £ ¡ #© £  X © # ¡ C ˆ £ ¡ #© £
H— – X –¼ '¤b§$4H¤© ¦„¤ $¤‰!§$4s§¢¤I2I¤‰!§$ ®
¤ ‘ H— – X –¼ §R§I)2¤© ɤ H— – X –¼ ¨6$ E ‰'6'¤)¦e A
X ˆ C # D ¥  \"£
X© #¡ C ˆ A ‘ P RECOND :
X ‘ X© #¡ C ˆ© ˆ # \" ©
§!6$ E §R§I)2¤6 } P$!¤¥ A
‘‘—© ¡ g X© #¡ C ˆ A ‘ ˆ a ¡ # £ 3© ˆ 
¦H¨§xiX –¼ §R§I)2¤© ɤ !6$ E b2‰§I§¡ B 1)\" Ë
# ˆ #©£ ‘‘— ¡ #  ¡  X© D C
‘xÎ2X µ ˆ4£§¡‰!©§£$ ® ¤ ‘ µ X&ÎI£¤¡‰!§$ ® ¤ ¦H¤bH ¤¤$ E 0'1 B – ¨6$ E ‰I6'§)¦ A X ˆ C # D ¥  \"£
¤ ‘—© ¡ g X © D C ˆ A ‘— ¡ #  ¡ 
H¨¤&'0'$@ B – X µ ¤© Ǥ H¤bH ¤¤$ E §0H¡ F – &¤© A 6¢# Ä X©G X Έ ˆ ©
‘ µ &e!R¤'C A
X Έ © # ¡
Figure 12.28
return # \" ©¥
$!¤
# \" ©¥R EMOVE -F LAW(
$!¤ ) // possibly updating
# 
1
) E FFECTS[
§¦§¢ §$Q© 8
© ¡¥£¡ ©£  ] = U PDATE(E FFECTS[ ],
©£ 
§$¨© 8
(the default) Æ ­h&$!!
\" g % # \" © ¥
D # ,
26¢H Å §1¨© 8
©£  static: , a plan, initially with just
$@
# 
# \" © ¥
1! ) returns an © ¡¥£¡
02§¢ function C ONTINUOUS -POP-AGENT(
Planning and Acting in the Real World 12. Chapter 30

13 UNCERTAINTY
function DT-AGENT( R§¦¤¢
© ¡¥£¡
) returns an $!¤
# \" ©¥
static: $Q0 !t§0
¡© © G¡  ¡
, probabilistic beliefs about the current state of the world
$!¤
# \" ©¥
, the agent’s action
update ¡© © G¡  ¡
$Q0 Ht§0
based on and # \" © ¥
$!! © ¡¥£¡
§¦§¢
calculate outcome probabilities for actions,
given action descriptions and current $Q0 HQ ¤2
¡© © G¡  ¡
select $!¤
# \" ©¥
with highest expected utility
given probabilities of outcomes and utility information
return $!¤
# \" ©¥
Figure 13.2
function E NUMERATE -J OINT-A SK( , e, P) returns a distribution overp p
inputs: , the query variable
p
e, observed values for variables E
P, a joint distribution on variables E ¨ ïi‰
Y /* Y = hidden variables */
Š ‹ ¨
Q pˆ % R‘
a distribution over , initially empty p
for each value of do ’ ¼ p
Q % R‘ ’ ¼ ˆ
E NUMERATE -J OINT( , e, Y, , P) ’ ¼ —–
return N ORMALIZE(Q ) ‘ ‹ˆ
function E NUMERATE -J OINT( , e, , §Rt$1d §1$d
 ¡ 3   £ 
, P) returns a real number
¼
if E MPTY ?( §1$d
£ 
) then return P e )¤R3 $$X X ¼ ˆ
‘ ¡   d
q %
F IRST( ) §$$d
£ 
return E NUMERATE -J OINT( , e, R EST(
r ), , P)
¼ £ 
§$1d —¡   Á
§¤R3 $$d “ ––
´
Figure 13.6
31

PROBABILISTIC
14 REASONING
function E NUMERATION -A SK( , e, ) returns a distribution over p # ¤ p
inputs: , the query variable
p
e, observed values for variables E
# ¤
, a Bayes net with variables E Y /* Y = hidden variables */ ¨ Ði‰
Š ‹ ¨
Q pˆ % R‘
a distribution over , initially empty p
for each value of do ’ ¼ p
extend e with value for ’ ¼ p
Q % R‘ ’ ¼ ˆ
E NUMERATE -A LL(VARS[ ], e) # ¤
return N ORMALIZE(Q ) ‘ ‹ˆ
function E NUMERATE -A LL( §$$d
£ 
, e) returns a real number
if E MPTY ?( ) then return 1.0 §1$d
£ 
s %
F IRST( ) §$$d
£ 
s
if has value in e 7
then return €¦‘ q „ xp uRÌ Á™– ˆ £
y ‘ ˆ wv t†
E NUMERATE -A LL(R EST( ), e) £ 
§$$d
else return y€‘¦‘ q '„ wp uÌ g– ˆ £ ´ r
ˆ v t † Á
E NUMERATE -A LL(R EST( ), e ) §$1d
£  ´
where e is e extended with
– w q ´
Figure 14.10
function E LIMINATION -A SK( , e, ) returns a distribution over
p # ¤ p
inputs: , the query variable
p
e, evidence specified as an event
# ¤
, a Bayesian network specifying joint distribution P ‘ o‹ X ™‹ ˆ
‘ X    Œ
% £  % £ \"© ¥ 
f§$$d — – `§$Q¤G
; R EVERSE(VARS[ ]) # ¤
£ 
§$$d
for each 1$d
£  in do
% £ \"©¥ 
– f§$¨¤¤G
M AKE -FACTOR e —£ \"© ¥  X £  d
§§1¨eG Á ‘ '1$sˆ
if 1$d
£ 
is a hidden variable then S UM -O UT( % £ \"© ¥ 
f§$Q¤G £ \"© ¥  £ 
§$Q¤G $1d
, )
return N ORMALIZE(P OINTWISE -P RODUCT( )) §£$\"Q©¤G
¥ 
Figure 14.12
32

33
# ¤
function P RIOR -S AMPLE( ) returns an event sampled from the prior specified by # §
# ¤
inputs: , a Bayesian network specifying joint distribution P ‘ o‹ X ™‹ ˆ
‘ X    Œ
%
x an event with elements v
for to do w z v
% ’ ¼
a random sample from P ’ ‹ˆ xp §RÌ Á
wv t† ‘ ˆ
¦‘ ’ ‹ „
return x
Figure 14.15
function R EJECTION -S AMPLING( , e, , ) returns an estimate of p # ¤ g ‘ Á ‹ˆ £
e
inputs: , the query variable
p
e, evidence specified as an event
, a Bayesian network # ¤
g
, the total number of samples to be generated
local variables: N, a vector of counts over , initially zero p
for = 1 to do
¶ ‚
%
x P RIOR -S AMPLE( ) # §
if x is consistent with e then
% k
N[ ] N[ ]+1 where is the value of k k p in x
return N ORMALIZE(N[ ]) p
Figure 14.17
function L IKELIHOOD -W EIGHTING( , e, , ) returns an estimate of p # ¤ g ‘ Á ‹ˆ £e
inputs: , the query variable
p
e, evidence specified as an event
, a Bayesian network # ¤
g
, the total number of samples to be generated
local variables: W, a vector of weighted counts over , initially zero p
for = 1 to do
¶ ‚
x, % —u
W EIGHTED -S AMPLE( ) # ¤
W W %$4k –
—
where is the value of u ’ —
U'4k – k p in x
return N ORMALIZE(W ) — p–
function W EIGHTED -S AMPLE( # ¤ , e) returns an event and a weight
%
x an event with elements; 1 v % —u
for = 1 to do v
if has a value in e
’ ‹ ’ ¼
then w ’o‹ ˆ £ ƒg—u
y u % ¦‘ o‹ „ xp §RÌ W¼
‘ ’ ˆ w v t † Á ’
else a random sample from P % ’ ¼ ‘ ¦‘ ’ ‹ '„ wvp uÌ Á ’ ‹ ˆ
ˆ t†
return x, u
Figure 14.19

34 Chapter 14. Probabilistic Reasoning
p— # ¤
function MCMC-A SK( , e, , ) returns an estimate of g e ‘ Á ‹ˆ £
local variables: N , a vector of counts over , initially zero
p– p
Z, the nonevidence variables in # §
x, the current state of the network, initially copied from e
initialize x with random values for the variables in Z
for = 1 to do
¶ ‚
N % —
14k –
N U'4k –
z ’ —
where is the value of in x k p
for each in Z do
„ ’
sample the value of in x from P „— ’ ‘ ¦‘ ’ „ 1Ñ Á ’ „ ˆ
ˆ given the values of ‘ ’ „ ˆ â†
µ in x
return N ORMALIZE(N ) p–
Figure 14.21

PROBABILISTIC
15 REASONING OVER TIME
function F ORWARD -BACKWARD(ev, $!§
£ \" £
) returns a vector of probability distributions
inputs: ev, a vector of evidence values for steps ¤)¤z
X  X w
$!§
£ \" £
, the prior distribution on the initial state, P X ‘ ‡ ˆ
local variables: fv, a vector of forward messages for steps X)¤X ˆ w
b, a representation of the backward message, initially all 1s
sv, a vector of smoothed estimates for steps w ¤ez
X   X
fv $!§„1— ˆ –
£ \" £ %
for w 1z% — w 
to do
ˆ Q¤•™  –
X—z ‘ H—  –
fv F ORWARD fv
– ev
for w$— w 
downto 1 do
% – ˆ y £—  – ‘
sv N ORMALIZE fv b
%
b BACKWARD b ev X ˆ ‘ H—  –
return sv
Figure 15.5
35

36 Chapter 15. Probabilistic Reasoning over Time
function F IXED -L AG -S MOOTHING( , , ) returns a distribution over X ‰ p p'D
T T a ‰
‘e
inputs: , the current evidence for time step
‰ p
, a hidden Markov model with p'D
T T transition matrix T y w± ±
, the length of the lag for smoothing a
©
static: , the current time, initially 1
f, a probability distribution, the forward message P , initially P RIOR ‘ ‰ ”$p ’‹ ˆ
“Œ Á‰ op'D –
— T T
B, the -step backward transformation matrix, initially the identity matrix
•
, double-ended list of evidence from
‰ P‰ p
“‘ e to , initially empty • ™w w
local variables: O ‰ X ©‰
O , diagonal matrices containing the sensor model information ‘e
add to the end of
‰p% ‰ P©‰ p
“‘ e
O ‰diagonal matrix containing P
˜
’‹ )p ˆ
‰ Á‰ ‘
if then
w •
f F ORWARD f % ‘ p X ˆ
‰
remove from the beginning of “2\"©‰ p
Œ e ‘ e ‰ P©‰ p
“‘ e
O diagonal matrix containing P
‘ \"©%e ‰ % ©‰ Á‹ \"©‰ p ˆ
‘e ‘e ‘
B O T BTO
Œ ce Œ ce ‰
%
©‰
‘e
else B BTO ‰
{ch©
z ’ © %
if • ˜ w then return N ORMALIZE f ˆ y ‘
B1 else return null
Figure 15.8
function PARTICLE -F ILTERING(e, , ) returns a set of samples for the next time step g §'a
#
inputs: e, the new incoming evidence
g
, the number of samples to be maintained
¤'a
#
, a DBN with prior P X , transition model P X X , and sensor model P E X ˆ ‘ ÁeŒ ˆ ‘ ‘ Œ $Œ ˆ
static: , a vector of samples of size , initially generated from P X
8 g‡ ‡ ˆ ‡ ‘
Á
local variables: , a vector of weights of size – g
for = 1 to do g
[ ] sample from P X X
8 % Á Œ ˆ ‡ w ‘— –
H y8
[ ] P eX – % w Œ Á ˆ 6—  v±
‘ –
% 8 W EIGHTED -S AMPLE -W ITH -R EPLACEMENT( , , g 8 – )
return 8
Figure 15.18

16 MAKING SIMPLE
DECISIONS
function I NFORMATION -G ATHERING -AGENT( 02§¢
© ¡¥£¡ ) returns an $!¤
# \" ©¥
static: , a decision network
5
integrate ¨§¦¤¢
© ¡¥£¡
into 5
% ¶ the value that maximizes — ™ ˜ Ä
‘ “ 9ˆ V® ˜ˆ©
‘ “ 9¤0$\" Ê
if — ˜ ˜ Ä
Б “ 9ˆ V® ˜ˆ©
‘ “ 9¤0$\" Ê
then return R EQUEST( ) ˜ “
else return the best action from 5
Figure 16.9
37

17 MAKING COMPLEX
DECISIONS
function VALUE -I TERATION( , ) returns a utility function §oT
a ™
§oT
a l
inputs: , an MDP with states , transition model , reward function , discount
8 B d
, the maximum error allowed in the utility of any state
™ Í | Í
local variables: , , vectors of utilities for states in , initially zero 8
e
, the maximum change in the utility of any state in an iteration
repeat
% ïÍ ; 0 | Í e %
for each state in 8  do
1§ – vÍ
% — | ’¢—§ –xB j
h†ƒ d
g 1 f
‡ ti„ – ¢6)0¦eQˆ
—| Í ‘| „ X † X „
if p§ – | Í Á
™ — Í e i©‚ e ˜ Á ¤„ –
—then % ™ —
x§ – | Í Á Á ¤„ – Í
—
until
k le d sˆ ™
™ z º ¦‘ d
return Í
Figure 17.5
38

39
function P OLICY-I TERATION( §oT
a
) returns a policy
§oT
a l
inputs: , an MDP with states , transition model8
Í | Í
local variables: , , vectors of utilities for states in , initially zero 8
m
, a policy vector indexed by state, initially random
repeat
% ïÍ
P OLICY-E VALUATION( , m Í , §oT
a )
W42I'$4¤I3
% r a¡ C #  D¥ #true
for each state in do 
if
j r ™t—)„ – ¢6‘)0¦eQ8„ ˆ j r f 1†ƒ
˜ | Í | „X †X ‡ t)„ – ¢6)0„ – m eQˆ
—| Í ‘| „ X— X „ then
”‘i ‚ | 22eQˆ j ”‚ 1§ m
iÇ% —
—| „– ¢ „ X † X „
Í r f 1†eTo –
‡ ƒ pn
”‚ P¢2'I$4§bI3
i % r a¡C # D¥ #
false
until ¢¦II14¤43
r a¡ C #  D¥ #
®
return
Figure 17.9

18 LEARNING FROM
OBSERVATIONS
function D ECISION -T REE -L EARNING( , © I$¤!'a ¤§!$ ¤Ã$e0¡
 3  G¡   £©©  ¡ T  k
, ) returns a decision tree
inputs: §(Ã10¡
 ¡ T  k
, set of examples
§Q§!$
  £©©
, set of attributes
© I$¤!'a
 3  G¡
, default value for the goal predicate
if §(RÃ$2¡
 ¡ T  k
is empty then return © I$¤!'a
 3  G¡
¤Ã$e0¡
 ¡ T  k
else if all have the same classification then return the classification
else if  £©©
§§1 is empty then return M AJORITY-VALUE( )  ¡ T  k
§(Ã10¡
else
% ©¡
R0¤2 C HOOSE -ATTRIBUTE( ,  ¡ T  k   £©©
§(Ã10¡ ¤§!$
)
`02!©
% ¡¡£ a new decision tree with root test 0§0
©¡
% {T
M AJORITY-VALUE( ) ’ §(RÃ$2¡
 ¡ T  k
for each value of ’ d
do 0§0
©¡
 ¡ T  k
% ’ §(RÃ$2¡ elements of
‰ with  ¡ T  k
§(Ã10¡ Š ’ œ w 2¤0 ©¡
f02§2
% ¡¡£© 3
D ECISION -T REE -L EARNING( £ ©
T ©0§¡0ý™P¤§!©$X ’ §¡(RÃ$2¡
 T  k
, )
add a branch to 02©
¡¡£
with label and subtree ’d 0¡2§R0
¡ £© 3
return 02©
¡¡£
Figure 18.6
40

41
function A DA B OOST( Ï F ¤Ã$2¡
 ¡ T  k
, , ) returns a weighted-majority hypothesis
inputs: §(Ã10¡
 ¡ T  k
, set of labelled examples g ˆ X  X Œ Œ
q\"¼ )¤)¤§‘ 4– X “¼ ˆ qu– X ‘
, a learning algorithm
F
Ï
, the number of hypotheses in the ensemble
local variables: w, a vector of example weights, initially g uiz
gº
h, a vector of hypotheses Ï
z, a vector of hypothesis weights Ï
for T = 1 to
do Ï
h (% —
e˜T – , w)  ¡ T  k
¤Ã$2¡ F
0%V$¦§§¡
£ \"££
for = 1 to do ¶ g
if h then uw ˆ—
Ñ “ ¼ ¨˜T – “ – ’ £ \"££¡ % £ \"££
å$¦§¤iV$¦§§¡ w — r–
for = 1 to do¶ g
if h then w“ – w ‘ “ ¼ ¨˜T –
w ˆ— % 1— r – e12§§{s4$$¦§§­t— r –
‘ £ \"££¡ ™ zˆ º £ \"££¡ s
w N ORMALIZE(w) %
z 12§§R¦$$¦§¤{vs˜yxv$˜T –
£ \" £ £ ¡ º ‘ £ \" £ £ ¡ ™ z ˆ p wu % —
return W EIGHTED -M AJORITY(h, z)
Figure 18.12
function D ECISION -L IST-L EARNING( §(RÃ$2¡
 ¡ T  k ) returns a decision list, or 2It6$¤G
¡ £ 3 
if §(RÃ$2¡
 ¡ T  k
is empty then return the trivial decision list \" ­g
% © a test that matches a nonempty subset of  ¡ T  k
§(Ã10¡ ¤Ã$e0¡
 ¡ T  k
§(RÃ$2¡
 ¡ T  k
‰
such that the members of are all positive or all negative ‰
if there is no such then return © 2It6$¤G
¡ £ 3 
if the examples in §(Ã10¡
 ¡ T  k
are positive then ‰ else % y\" s  ¤¡ \" g %
­zy\"
return a decision list with initial test and outcome and remaining tests given by © \"
D ECISION -L IST-L EARNING( ) ¡
™Ð¤ ÃT$2¡
k ¡  k
¤ ÃT$2¡
‰
Figure 18.17

19 KNOWLEDGE IN
LEARNING
function C URRENT-B EST-L EARNING( §(RÃ$2¡
 ¡ T  k ) returns a hypothesis
% } any hypothesis consistent with the first example in ¤Ã$e0¡
 ¡ T  k
for each remaining example in do ¤Ã$2¡
 ¡ T  k
¡ }
if is false positive for then
% }
choose a specialization of consistent with
} ¤Ã$e0¡
 ¡ T  k
¡ }
else if is false negative for then
% }
choose a generalization of consistent with
} ¤Ã$e0¡
 ¡ T  k
if no consistent specialization/generalization can be found then fail
}
return
Figure 19.3
function V ERSION -S PACE -L EARNING( ) returns a version space §(RÃ$2¡
 ¡ T  k
local variables: , the version space: the set of all hypotheses
—
— %
the set of all hypotheses
for each example in ¡ §(Ã10¡
 ¡ T  k do
if is not empty then
— % — V ERSION -S PACE -U PDATE( , ) — ¡
return —
function V ERSION -S PACE -U PDATE( , ) returns an updated version space
— ¡
— % ” ‰  —
|{» ” is consistent with Š ¡
Figure 19.5
42

43
n
function M INIMAL -C ONSISTENT-D ET( , ) returns a set of attributes A
n
inputs: , a set of examples
, a set of attributes, of size
A #
for % 4 ˆ # X   
g†X
do
for each subset of of size do
’ vA A n
if C ONSISTENT-D ET ?( , ) then return ’ A ’ A
n
function C ONSISTENT-D ET ?( , ) returns a truth-value A
inputs: , a set of attributes
An
, a set of examples
}
local variables: , a hash table
¡ n
for each example in do
} ¡
if some example in has the same values as for the attributes A
but a different classification then return ¡¤t$¤G
 
¡ } ¡
store the class of in , indexed by the values for attributes of the example A
return R§©
¡ 3£
Figure 19.11

44 Chapter 19. Knowledge in Learning
¤'Q1¨© ¤Ã$2¡
© ¡ C£   ¡ T  k
function F OIL( , ) returns a set of Horn clauses
inputs: §(Ã10¡
 ¡ T  k
, set of examples
§I$¨©
©¡ C£ 
, a literal for the goal predicate
local variables: ¤¤41¤¥
 ¡  3 
, set of clauses, initially empty
while ¤Ã$e0¡
 ¡ T  k
contains positive examples do
% ¡  3 
P¤41¤¥
N EW-C LAUSE( , ) © ¡ C£   ¡ T  k
§I$Q© ¤Ã$2¡
remove examples covered by from ¤ ÃT$ek0¡
¡ I$@¥
¡  3 
add I$@¥
¡  3 
to ¤¤41¤¥
 ¡  3 
return §I$@¥
 ¡  3 
function N EW-C LAUSE( , §I$Q© §(Ã10¡
© ¡ C£   ¡ T  k
) returns a Horn clause
local variables: ¤I$Q©
©¡ C£ 
, a clause with ¤41¤¥
¡  3 
as head and an empty body

, a literal to be added to the clause
§(Ã10¡ 2b¤¨$2¡
 ¡ T  k a ¡ a # ¡© k
, a set of examples with values for new variables
¤Ã$2ˆf¤Ã$e0¡ 2'¤10¡
 ¡ T  k ¡ %  ¡ T  k a ¡ a # ¡ © k
while ¤Ã$2¡ 2'b§10¡
 ¡ T  k a ¡ a # ¡ © k
contains negative examples do
¡I3$@¥
 C HOOSE -L ITERAL(N EW-L ITERALS( ), %
e )  ¡ T  k a ¡ a # ¡ © k
¤Ã$e0¡ ¦'b§$2¡
append to the body of ¡I$@¥
 3  
`§(Ã10¡ 2b¤¨$2¡
%  ¡ T  k a ¡ a # ¡© k
set of examples created by applying E XTEND -E XAMPLE
§(R ÃT$k2¡ a¦¡'b§$2¡
¡
to each example in a #¡© k
return I$@¥
¡  3 
function E XTEND -E XAMPLE( , $¦¤¨6t (Ã10¡
  £ ¡© ¡ T  k
) returns
if (RÃ$2¡
¡ T  k
satisfies $¦§6t
 £¡©
then return the set of examples created by extending Ã$2¡
¡ T  k with
each possible constant value for each new variable in 12§6 
 £¡©
else return the empty set
Figure 19.16

20 STATISTICAL LEARNING
METHODS
function P ERCEPTRON -L EARNING( 0$!¤‰# ¤Ã$e0¡
9 £ \" u© ¡  ¡ T  k
, ) returns a perceptron hypothesis
inputs:  ¡ T  k
§(Ã10¡, a set of examples, each with input x ‘ §¤)X Œ k w
kX  
and output 7
0$!¤‰#
9£ \" u© ¡
, a perceptron with weights X “ ·, and activation function
r w ˆ  
v ¤ C
repeat
for each in do ¤Ã$e0¡
 ¡ T  k ¡
‘
— ¡ – “ k “ ·
‘ # ˆ C ™
‡ ~“ 7 r % &6#
%
}
“6s‰¬p— p – h§£ n £
—¡ – “ kžy€‘“6sˆ | Cžy£§£ n y š ’ “ · % “ ·
#
until some stopping criterion is satisfied
return N EURAL -N ET-H YPOTHESIS( 0$§‰#
9£ \" u©¡ )
Figure 20.22
45

46 Chapter 20. Statistical Learning Methods
function BACK -P ROP -L EARNING( 0$!¤‰# ¤Ã$e0¡
9 £ \" u© ¡  ¡ T  k
, ) returns a neural network
inputs: §(Ã10¡
 ¡ T  k
, a set of examples, each with input vector x and output vector y
0$!¤‰#
9£ \" u© ¡
, a multilayer network with layers, weights
F , activation function
³’ “ · C
repeat
for each in¡ ¤Ã$e0¡
 ¡ T  k
do
for each node in the input layer do
r ¤¡ “ ~% “ 
— – k
for = 2 to do Ï 
 ³’ “ · % ’ 6
#
“ “ r
‘ ’ 6sb~% ’ 
# ˆ C
for each node in the output layer do
# C
‘ ’ gp¤¡ – ’ – ɂ‘ ’ Hsˆ | Ç% ’ m
† ™ — ˆ y
for = to 1 do z ™ Ï 
for each node in layer do r# C
m  m
’ ³’ “ · ’ r ‘ “ 6sˆ | ~% “
z{’ 
for each node in layer
 do
m
“ Éy š ’ ³’ “ · % ³’ “ ·
† y ’
until some stopping criterion is satisfied
return N EURAL -N ET-H YPOTHESIS( 0$§‰#
9£ \" u©¡ )
Figure 20.27

REINFORCEMENT
21 LEARNING
function PASSIVE -ADP-AGENT( §¦§¢
© ¡¥£¡
) returns an action
inputs: R§¦¤¢
© ¡¥£¡
, a percept indicating the current state and reward signal |  | £
m
static: , a fixed policy
§oT
a
, an MDP with model , rewards , discount
j ƒ d
Í
, a table of utilities, initially empty
( X gg
, a table of frequencies for state-action pairs, initially zero
, a table of frequencies for state-action-state triples, initially zero
i X ( X 
, , the previous state and action, initially null
if | 
is new then do [ ] ; [ ]Í |
B $h% 
| £ | £
$h% 
|
if 
is not null then do
  (X g
increment [ , ] and
©
iX(X g
[ , , ]
©   ( g
|   
for each such that [ , , ] is nonzero do
iX X g
[ , , ]   l  ( X g ©   ( h% ©
[ , , ]/ [ , ] 
% ÐÍ 0aoT Í m
iX X
VALUE -D ETERMINATION( , , )
if T ERMINAL ?[ ] then , |  %
null else , w  ,  %
| gv  m —| –
return 
Figure 21.3
47

48 Chapter 21. Reinforcement Learning
function PASSIVE -TD-AGENT( R§¦¤¢
© ¡¥£¡
) returns an action
inputs: R§¦¤¢
© ¡¥£¡
, a percept indicating the current state and reward signal |  | £
m
static: , a fixed policy
Í
, a table of utilities, initially empty
‚ g
, a table of frequencies for states, initially zero
£  
, , , the previous state, action, and reward, initially null
if | 
is new then [ ] Í | h% | 
£
if 
is not null then do
increment [ ]  ‚ g
[ ] Í %  ­§ – Í
’ — ås¤6§ – ‚ ˆ š
’ £ˆ ‘— g d H§ – Í p— |  – Í
‘— ™
if T ERMINAL ?[ ] then , , |  % £   null else , , h£  
|  % , m | —|
$£ @ –
,
return 
Figure 21.6
function Q-L EARNING -AGENT( ) returns an action 02§¢
© ¡¥£¡
inputs: R§¦¤¢
© ¡¥£¡
, a percept indicating the current state and reward signal |  | £
static: , a table of action values index by state and action
„
(  X g
, a table of frequencies for state-action pairs
£
, , , the previous state, action, and reward, initially null

if is not null then do
increment [ , ]   (X g
™x— | ¤X | † –p„ f 1 gƒ d ’Âs¤‘6eiX$ – ( X ˆ š ’¢§§R p„ 1§§R p„
‡ £ˆ — g —X – % —X – H§§R p„
‘—  X –
i %£   | 
if T ERMINAL ?[ ] then , , null
| £ H— | )X | † – ( X bXQ— | X | † –p„ ˆ ‚ f ‡$geTo | h£  
else , , , ‘  g  i ƒpn  %  ,
return
Figure 21.11

22 COMMUNICATION
function NAIVE -C OMMUNICATING -AGENT( ¨§¦¤¢
© ¡¥£¡
) returns 1!
# \" © ¥
static: , a knowledge base
E ¢¡
$Q0
¡© ©
, the current state of the environment
#$\"!¤
©¥
, the most recent action, initially none
P1¨0
% ¡© ©
U PDATE -S TATE( , © ¡¥£¡ # \" ©¥ ¡© ©
02§¢ $!¤ 1¨0
, )
`$¦$u
%  a£ \"
S PEECH -PART( ) ¨§¦¤¢
© ¡¥£¡
%`¤!©R1o¤
¥ #  T¡
D ISAMBIGUATE(P RAGMATICS(S EMANTICS(PARSE( )))) $¦$u
 a£ \"
¡‰#$\"­g $¦$u
if = a£ \" and $!e
# \" © ¥
is not a S AY then /* Describe the state */
return S AY(G ENERATE -D ESCRIPTION( )) ¨$¨2
¡© ©
else if T YPE[ ¤!R1o¤
¥ © #  T¡
] = Command then /* Obey the command */
return C ONTENTS[ ¤!R1o¤
¥ © #  T¡
]
¤!R1o¤
¥ © #  T¡
else if T YPE[ ] = Question then /* Answer the question */
E ¢¡ V¤y0R1
% £¡ u #
A SK( , ¥ © #  T¡
¤!R$o§
)
return S AY(G ENERATE -D ESCRIPTION( )) ¤y0R1
£¡ u #
else if T YPE[ ¤!R1o¤
¥ © #  T¡
] = Statement then /* Believe the statement */
T ELL( , C ONTENTS[
E x¡ ]) ¥ © #  T¡
¤!R$o§
/* If we fall through to here, do a “regular” action */
return F IRST(P LANNER( , E))
¢¡ ¡© ©
$Q0
Figure 22.3
49

50 Chapter 22. Communication
function C HART-PARSE( £  T T £  a£ \"
$op$¦eC $¦$u , ) returns ©£  D
§$4§¥
§14¤¥
% ©£  D array[0. . . L ENGTH( )] of empty lists $¦$u
 a£ \"
A DD -E DGE( | ± X ˆ X ˆ–
) 3 — ±
for % 4
from 0 to L ENGTH( ) do  a£ \"
$¦$u
S CANNER( , [ ]) 121yu
 a£ \"
return §$I¤¥
©£  D
procedure A DD -E DGE( ) ')2¡
¡ C a
/* Add edge to chart, and see if it extends or predicts another edge. */
if ¡ C a
I)¦¡
not in [E ND( ©£  D
§$4§¥ )] then ¡ C a
')2¡
append to [E ND( ¡ C a
')2¡)] ©£  D
§$I¤¥ ¡ C a
I2¡
if ¡ C a
')2¡
has nothing after the dot then E XTENDER( ) ¡ C a
I)2¡
else P REDICTOR( ) I)¦¡
¡ C a
procedure S CANNER( , ) a¦$u ¶
£ \"
/* For each edge expecting a word of this category here, extend the edge. */
for each A X r X – in 3 [ ] do š — › E ©£  D
¶ §$4§¥
if ¦$u
a£ \"
is of category then
A DD -E DGE( 3 Î X zE
hVep’ r X  –
) µ š — ›
procedure P REDICTOR( )3 A X ¶ 0–
X š — •E
›
/* Add to chart any rules for that could help extend this edge */ E
for each Eˆ 3
in R EWRITES -F OR( , ‘ d ) do E £  T T £
1op$¦eC
A DD -E DGE( ) E X r X r– 3 d —
procedure E XTENDER( E X9§X –¶
) 3 d — I
/* See what edges can be extended by this edge */
† )p %
the edge that is the input to this procedure
for each 3 Î
‡`X r X  – in [ ] do š — › | µ ©£  D
¶ §$4¤¥
if µ then | µ w
A DD -E DGE( 3 Î Xˆ
hïYX  –
) † Yš
p — ›
Figure 22.9

23 PROBABILISTIC
LANGUAGE PROCESSING
® $0¨©
© k¡
function V ITERBI -S EGMENTATION( , ) returns best words and their probabilities
inputs: 10©
© k¡
, a string of characters with spaces removed
®
, a unigram probability distribution over words
L ENGTH(% &# ) $0¨©
© k¡
`$¦$u
%  a£ \"
empty vector of length v z {’
% ©¡
R2¤0
vector of length , initially all 0.0 v z {’
%
[0] 1.02¤0
©¡
/* Fill in the vectors best, words via dynamic programming */
for = 0 to do v
for = 0 to do z •™  ¶
a \"
[ :]
¦£$
 u r $2hv¦$—uu
© k¡© % a£ 
%
\"
L ENGTH( )
if [ ] 0¤¡0 y 21yu ®
© [ - ] a£ \" [ ] then ©¡
0§0  u
[]a¦$\"u ® %  0§0
£ [ ] [ ©¡ ] 2¤0ýy
©¡ ‰ Š™ 
[]¦$g%  $¦$u
a£ \" u  a£ \"
/* Now recover the sequence of best words */
sequence the empty list %
 % v
while do  ˆ ˜
$a¦£$\"u §b§R2
¡¥ #¡ 3 S¡
push [ ] onto front of 
 %
L ENGTH( ™ [ ])  a£ \"
$¦$u 
/* Return sequence of best words and overall probability of sequence */
return , 0¤0 0b¤R¤2
©¡ ¡¥ #¡ 3S¡
[] 
Figure 23.2
51

24 PERCEPTION
function A LIGN( ¤')oT I)o6
¡ a \"
, ¡ C  T
) returns a solution or failure
inputs: IoH
¡ C  T
, a list of image feature points
¤')oT
¡ a \"
, a list of model feature points
for each Ì Ì
 bX Ž X Œ Ì
X Y X ¦
in T RIPLETS( ) do
§')oT
¡ a \"
¡ C  T
')o6
for each Ž Œ in T RIPLETS(
 ‚ ) do
„ %
F IND -T RANSFORM( X ŽY X Œ¦  bX Ž bX Œ Ì
, Ì Ì  ‚ )
if projection according to explains image then
„
return „
return failure
Figure 24.22
52

25 ROBOTICS
function M ONTE -C ARLO -L OCALIZATION( , , , , ) returns a set of samples °  g  ¡ a \"
§oT §')oT

inputs: , the previous robot motion command
z, a range scan with readings Ï Y0X Œ °
‹°X
g
, the number of samples to be maintained
¤')oT
¡ a \"
, a probabilistic environment model with pose prior P X , ˆ ‘ ‡
motion model P X X , and range sensor noise model
Á $Œ ˆ Î
‡ 2X ‡ ‘ ‘vbÁ ”ˆ £
Œ Œ
§oT

, a 2D map of the environment
static: , a vector of samples of size , initially generated from P X
8 g ˆ ‡ ‘
local variables: , a vector of weights of size – g
for = 1 to do g
[ ] sample from P X X
8 % ÁeŒ ˆ ‡ w Î X— –
`Q y8 ‡ w ‘ I
[ ] 1
– %
for =1 to do ¶ Ï
% y°
E XACT-R ANGE( , [ ], 
§oT 8 ¶ )
 Ž •–
s —– ‘° w YÁ “ ° w ”ˆ £
Œ
% 8
– 1Q –
% —
 Œ g
W EIGHTED -S AMPLE -W ITH -R EPLACEMENT( , , 8 – )
return 8
Figure 25.8
53

26 PHILOSOPHICAL
FOUNDATIONS
54

27 AI: PRESENT AND
FUTURE
55