Given :

StSpace :
positions ,
x =

20 ,
1 ,
2
,
3
, . . .

., 2003
Total of 100 + 1 discrete locations
as the
given ID path is 100 meters
long
State Transition Model
Mon or :

(doesn't move)

[0
0 .
09 k =
;
PEKIF1)
Kimos a
=
,

Mode x) 201 Correcta =

m

Algorithm :
ID Bayes Filte (4t Et) ,

for all do :
up
be (x+ )
1 bet(u ) = +
[P((t -
) - -

bel (e)
2 .

bel(xz) =
n .

PEt(e) ·

end far
return bel (at)

#to
doesn't -
know abou to se

So, It'sBelief can be considered to be

uniformly distributed
all discrete
across 101
positions
bil(40) =
bel(nd) =

L-)
 TIMESTEP Y =
1

-
Robot moves Im alead.

&
rediction :
bel( ) ,
= P( (4) be ,

0Pool
#0 =

,
bill =

+ 0 + 0 ....

]
=
u

Robot can't
move backwards

# : Sel(x 1) ,
=

= =(4) be
= [p(x ,
=
1(x 0) =
+ p(x 1(xo ,
=
=

1)]
! (970s
=
# 2 : bel (n =

2) = P( (40) ,
- be (

= (p(x 2(40 0)
= =

+
p(x , 2(40 1) = =

+
p(x) 2(xo 2)]
= =

00
%05
09 +
+
=

=
101
 = 3 : ben, 3) =

= P( =

3) be .

= [P(x 3(x. 1)
= =
,

+
p(x, 3(x 2)
=
=

+ p(a
,
3/40 =
=

3)]
=

Ta
Similarly
,
bel(x i) = ,
= is

This is because ,
before receiving first sensor reading
,
the robot isn't aware
of its location and
according to

state transition model the model has to take

, any
passage of
the three actions with the time As ,
of .

motion model will influence

timestep increases by 1 ,

The robot would

the
belief distribution .
most
likely
to Im than at the same position more
more
stay or

I m ahead .
Due to this
,
it is less likely for the robot

position chances
to be in 2 = 0
,
as
for moving
backward is counted as zero .
To be at X = 1,

it's influenced by X = 0
.
To be at X = 2 ,
it's influenced
by X =
1 and X = 0 .
And for all positions till x = 100
,

all
influenced by their last
preceding positions
.
are 2

Due to this ,
we can observe

Li
=
Correction :
After receiving first sensor
reading : z
,
=

U1
= 1

P(z (x ) , ,
=

[09,MFZ
bel() =
7 .
P(z( ) ,
·
bet (x)

bel(x 0) 4
,
= =
x p(7 ( 0), ,
=
x
bel (x , =
0)

Sell =
) 0
bel(u 2) = =

4x0
.

1x
3
,

: go value
!
bel (x, =
100) =
nx0 .

1 x

We know
, bel()
=>
y 20005095 99x +

>
- (0 . 005 + 0 855
.
+ 9 .
9) =
102

-
 belo t
bel (x ,
=
2) =
n x
0 .
1 +01 =
5538 = 0 . 0092

bel(x ,
=

3) = x 0 9 .

xY01 =538 = 0 .

0072

I i
100) 1xY01 = 0092
(x
0
bel ,
= =
y>
0 .
538 = .

>
-
belf) =

[18:0 ..... 3 un

~ 7 9 .

% belief at X =
1 for t =
!
 -

AT TIMESTEP t = 2
-

Robot Im ahead
Assuming
mors

Prediction :

bit() = p(x() -

be (x)

F
= 0 : Jel (2 0)= =
p(x 0(x 0) ,
=

,
=
.
bel (x ,
=
0)
= 0 .
05
x52
5040 = 0 000023
=
.

For 1 : bil (n =
1) =
p(x 1(x 0) bel(x 0)
=

,
=
·

,
=

+
p(x, 1(x 1) bel(x 1)
=

,
=
.

,
=

= 0 .

9 + 0 .

05
= (0 . 9 + 0 .

5)

=0
0 00439 ...
.
Exc = 2 :

Jil(x =
=
2) =
p(xz 2(x 0) bel (n 0)
=

,
=
·

,
=

+
p(x 3(x =
=

,
=
1) be (n .
,
=
1)
>
- p( =
=
2(n ,
=
2) bel (n
·

,
=
2)

- -

=3 :

Jil(x, 3) = =

p(xj 3(x 2) =

,
=
.
bel(x 1) ,
=

+ P(4y =
3(x ,
=
2) 2) ·

bel(x ,
=

+ p(xz 3(x =

,
=
3) bel(x 3) ·

,
=

>175
-

538

=> 0 .
0220

Fo4
: Bel (2 =
4) =
p(x= 4)x 2) bel(x 2)
=

,
=
-

,
=

+ P(x 4(x 3) bel(x 3)

=

,
=
-

,
=

+ p(xz 4(x 4) be (x, 4)

=

,
=
.
=

= (0 . 05 + 0 9 + 0
. .

05) =538

Sinikely for all x34 ,

Jel (mm) =
5/30

Gano
o
-
be (1) =
Correction
-
After Sensor reading ,
zz =
2

Je (2)
bel(x) =
1 .

p(Eck) .

Let plzzIk)) =

CoyEEUA
(0) n(0 /3040)
Then ,

. (0 1)
bel(uc 0 1 =
= = ·

be(nz 1) m(0
= = .

2)() =

(18 /93090)
.

bel (mz 7 7(0

= = .

9)(9) n(2789 /3040) = .

bel(n = 3) =(0 1)(40) 7(55 /3090) .

= .

2 (2 4)= =

y(0 1) (0) .

;; " Get value

bel(x = 100) y (0 1) (i)

= .

48 . 5

a
We know ,

bu
- 1997 43040
255 .

E
-

=
n
0 1
.
+ 10 .

9 + 2789 1 . + 55 1. + 38003 = 43040

- n =
3092
bel (x =
=
0) = 0 .

2/6793 .
2 =67932

bel (n, 1) =
= 18 .

9/6743 . 2
= 18
% 7932
bel (x z
=
2) = 2789 .
9/6743 .
2
=

2789/67432
bel(x, 3) = = 55 .

9/6743 .
2
=
55//67932
bel(x = 4)
4 400/7432
=
=

-
↑

(ic %7932
&

bel =
100) =
9

bel
(2) =

Gina 2
x = 2
1323
.....
 TIMESTEP t > =

Robot Im alread
.
Assuming
moves

Prediction :
5) = ps) bed (

the sunuation
After calailating using Python ,
for all 101 positions;

bet (ns) =

20 .
0
,
0 . 0002 ,

0059 ,
0 0232
.

0 059 ,
,
0 3728
-
.

0006 ,
0 .
.

...

0 .

00593

correction :
Sensor
Reading E = 4

bel (ns) =

y +
P(zy(s) + bet (x)

After doing calculations

the
using Python
,
we
get ,

bel (13) =

90 .
0
,

0 2009 ,
.
0 .
0002
,

0 005
.
0 . 019

,
,

0
059,
. 0049 0 .
0049
,
,
-

....,
0 .

00493
 The belief at
position x
=
3 remains
higher even
after
the sensor detects the position 4 ,
because
of the robot's
motion model .
The robot's movement probabilities cause some

belief to
spill over to
nearby positions ,
Like =3 .
So,
,
even

though the sensor

confirms position . The robot still
4
=
x

maintains some
belief at position 3 due to its
previous
movement history
.
 Sp t=

doesn't
Assuming robot
move-stays at x =
4

Mediction :
5) = ps) bel l

the sunuation
After calailating using Python ,
for all 101 positions;

E 0324 , 0 2863 0 0151 0 0049 ,

bel(nz)
0 0 2035 ,
0019
.

0 0 0 0 0 ,
.
.

=
.
.
. ,
,
.
,
.

0 .
0049 ,
0 . 0049; ...,
0 .

0049]
correction Sensor
Reading =4
:

bel
(nq) =

y +
P(z , (4) +
bet(x)
After doing calculations
the
using Python
,

90 0 7842 0 0619
0. 0046 ,
bel (13) 0099
0 0003 0
.

0, 0 0 , ,
.
.

= .
. .

, , ,

. 0015,
0 0 0015,
.

,..., 0 .

00157