Computers and Mathematics with Applications 63 (2012) 133143

Contents lists available at SciVerse ScienceDirect

Computers and Mathematics with Applications

journal homepage: www.elsevier.com/locate/camwa

A numerical method for solving m-dimensional stochastic ItVolterra

integral equations by stochastic operational matrix
K. Maleknejad , M. Khodabin, M. Rostami
Department of Mathematics, Karaj Branch, Islamic Azad University, Karaj, Iran

article info abstract

Article history: The multidimensional ItVolterra integral equations arise in many problems such as an
Received 12 April 2011 exponential population growth model with several independent white noise sources. In
Received in revised form 29 October 2011 this paper, we obtain a stochastic operational matrix of block pulse functions on interval
Accepted 31 October 2011
[0, 1) to solve m-dimensional stochastic ItVolterra integral equations. By using block
pulse functions and their stochastic operational matrix of integration, m-dimensional
Keywords:
stochastic ItVolterra integral equations can be reduced to a linear lower triangular
Block pulse functions
Stochastic operational matrix
system which can be directly solved by forward substitution. We prove that the rate of
Stochastic ItVolterra integral equations convergence is O(h). Furthermore, a 95% confidence interval of the errors mean is made,
It integral the results shows that the approximate solutions have a credible degree of accuracy.
Brownian motion process 2011 Elsevier Ltd. All rights reserved.

1. Introduction

We know that stochastic ItVolterra integral equations arise in many problems in mechanics, finance, biology, medical,
social sciences, etc. So the study of such problems is very useful in application and there is an increasing demand for
studying the behavior of a number of sophisticated dynamical systems in physical, medical and social sciences, as well
as in engineering and finance. These systems are often dependent on a noise source, on a Gaussian white noise, for example,
governed by certain probability laws, so that modeling such phenomena naturally requires the use of various stochastic
differential equations [15] or, in more complicated cases, stochastic ItVolterra and ItVolterraFredholm integral
equations and stochastic integro-differential equations [610]. Because in many problems such equations of course cannot
be solved explicitly, it is important to find their approximate solutions by using some numerical methods [14,810].
In recent years, orthogonal functions or polynomials, such as block pulse functions, Walsh functions, Fourier series,
Legendre polynomials, Chebyshev polynomials and Laguerre polynomials, were used to estimate solutions of some systems
such as integral equations, [714]. In this paper we use of block pulse functions and stochastic integration operational matrix.
We consider the following m-dimensional linear stochastic ItVolterra integral equation,
t m
t
X (t ) = f (t ) + (s, t )X (s)ds + j (s, t )X (s)dBj (s) t [0, T ),
0 j =1 0

where X (t ), f (t ), (s, t ) and j (s, t ), j = 1, 2, . . . , m, for s, t [0, T ), are the stochastic processes defined on the same
probability space (, z, P ), and X (t ) is unknown. Also B(t ) = (B1 (t ), B2 (t ), . . . , Bm (t )) is an m-dimensional Brownian
t
motion process and 0 j (s, t )X (s)dBj (s), j = 1, 2, . . . , m, are the It integrals.
This paper is organized as follows.

Corresponding author. Tel.: +98 21 732 254 16; fax: +98 21 732 284 16.
E-mail addresses: maleknejad@iust.ac.ir (K. Maleknejad), m-khodabin@kiau.ac.ir (M. Khodabin), m.rostami@kiau.ac.ir (M. Rostami).

0898-1221/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.camwa.2011.10.079
134 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143

In Section 2, we describe the basic properties of the block pulse functions and functions approximation by block pulse
functions and integration operational matrix. In Section 3, we obtain the stochastic integration operational matrix. In
Section 4, we solve stochastic ItVolterra integral equations with several independent white noise sources by using a
stochastic integration operational matrix. Section 5 is devoted to error analysis proposed method and in Section 6, we report
our numerical finding and demonstrate the accuracy of the proposed scheme by considering numerical examples. Finally,
Section 7 gives some brief conclusions.

2. Block pulse functions (BPFs)

BPFs have been studied by many authors and applied for solving different problems; for example, see [714]. The goal
of this section is to recall notations and definition of the block pulse functions, state some known results, and derive useful
formulas that are important for this paper. These have discussed thoroughly in [15,16].

2.1. Definition of BPFs

We define the m-set of BPFs as

(i 1)h t < ih,

1
i (t ) = (1)
0 otherwise

with t [0, T ), i = 1, 2, . . . , m and h = m

T
.
The elementary properties of BPFs are as follows
(1) Disjointness: The BPFs are disjointed with each other in the interval t [0, T )
i (t )j (t ) = ij i (t ), (2)
where i, j = 1, 2, . . . , m and ij is Kronecker delta.
(2) Orthogonality: The BPFs are orthogonal with each other in the interval t [0, T )
T
i (t )j (t )dt = hij , i, j = 1, 2, . . . , m. (3)
0

(3) Completeness: If m , then the BPFs set is complete; i.e. for every f L2 ([0, T )), Parsevals identity holds,
T

f 2 (t )dt = fi2 i (t )2 , (4)
0 i=1

where
T
1
fi = f (t )i (t )dt . (5)
h 0

Vector form: Consider the first m terms of BPFs and write them concisely as m-vector
T
(t ) = 1 (t ), 2 (t ), . . . , m (t ) , t [0, T ).

The above representation and disjointness property follows

1 (t ) 0 0

0 2 (t ) 0
(t ) T (t ) = ,

.. .. ..
.
.. (6)
. . .

0 0 m (t ) mm
furthermore, we have

T (t ) (t ) = 1,
and,

(t ) T (t )F T = DF (t ), (7)
T
where DF usually denotes a diagonal matrix whose diagonal entries are related to a constant vector F = f1 , f2 , . . . , fm

.
 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143 135

2.2. Functions approximation

An arbitrary real bounded function f (t ), which is square integrable in the interval t [0, T ), can be expanded into a
block pulse series in the sense of minimizing the mean square error between f (t ) and its approximation
m

f (t ) fm (t ) = fi i (t ), (8)
i=1

where fi is the block pulse coefficient with respect to the ith BPF i (t ). In the vector form we have,

f (t ) fm (t ) = F T (t ) = T (t )F , (9)

where
T
F = f1 , f2 , . . . , fm .

Let k(s, t ) L2 [0, T1 ) [0, T2 ) . It can be similarly expanded with respect to BPFs such as

k(s, t ) = T (s)K (t ) = T (t )K T (s), (10)

where (s) and (t ) are m1 and m2 dimensional BPFs vectors respectively, and K = kij , i = 1, 2, . . . , m1 , j =

1, 2, . . . , m2 is the m1 m2 block pulse coefficient matrix with
T1 T2
1
kij = k(s, t )i (s)j (t )dtds,
h1 h2 0 0

where h1 =
T1
m1
, h2 = T2
m2
. For convenience, we put m1 = m2 = m.

2.3. Integration operational matrix

t
Computing i (s)ds follows
0

0 0 t < (i 1)h,

t
i (s)ds = t (i 1)h (i 1)h t < ih, (11)
ih t < T .
0
h

Since t (i 1)h, equals to 2h , at the mid-point of [(i 1)h, ih), we can approximate t (i 1)h, for (i 1)h t < ih, by 2h .
t
Now expressing 0
i (s)ds, in terms of the BPFs follows

t
h
i (s)ds 0, . . . , 0, , h, . . . , h (t ), (12)
0 2

where 2h is the ith component of vector.

Therefore, [15],
t
(s)ds P (t ), (13)
0

where the operational matrix of integration is given by

1 2 2 2

0
1 2 2

h 0 0 1 2 .

P = (14)
2. .. .. .. ..

.. . . . .


0 0 0 1 mm

So, the integral of every function f (t ) can be approximated as follows

t t
f (s)ds F T (s)ds F T P (t ). (15)
0 0
136 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143

3. Stochastic integration operational matrix

The It integral of each single BPFs i (t ) can be computed as follows,

0 0 t < (i 1)h,

t
i (s)dB(s) = B(t ) B((i 1)h) (i 1)h t < ih, (16)
0 B(ih) B((i 1)h) ih t < T .

Since B(t ) B((i 1)h) is equal to B((i 0.5)h) B((i 1)h), at the mid-point of [(i 1)h, ih), we can approximate
B(t ) B((i 1)h), for (i 1)h t < ih, by B((i 0.5)h) B((i 1)h).
t
Now expressing 0 i (s)dB(s), in terms of the BPFs follows
t m

i (s)dB(s) B((i 0.5)h) B((i 1)h) i (t ) + B(ih) B((i 1)h) j (t ), (17)
0 j=i+1

and it has the vector form,

t
i (s)dB(s) 0, . . . , 0, B((i 0.5)h) B((i 1)h), B(ih) B((i 1)h), . . . , B(ih) B((i 1)h) (t ), (18)
0

in which the ith component is B((i 0.5)h) B((i 1)h).

Therefore
t
(s)dB(s) PS (t ), (19)
0

where the stochastic operational matrix of integration is given by

h
B B(h) B(h) B(h)
2
3h
0 B B(h) B(2h) B(h) B(2h) B(h)

2


5h
.

PS =
0 0 B B(2h) B(3h) B(2h) (20)

2


. .. .. .. ..
.

. . . . .


(2m 1)h

0 0 0 B B((m 1)h)
2 mm

So, the It integral of every function f (t ) can be approximated as follows

t t
f (s)dB(s) F T (s)dB(s) F T PS (t ). (21)
0 0

4. Solving m-dimensional stochastic ItVolterra integral equations by stochastic operational matrix

We consider following linear stochastic ItVolterra integral equation with several independent white noise sources,
t m
t
X (t ) = f (t ) + (s, t )X (s)ds + j (s, t )X (s)dBj (s) t [0, T ). (22)
0 j =1 0

Our problem is to determine the block pulse coefficient of X (t ), where X (t ), f (t ), (s, t ) and j (s, t ), j = 1, 2, . . . , m,
for s, t [0, T ), are the stochastic processes defined on the same probability space ( , z, P ). Also B(t ) =
t
(B1 (t ), B2 (t ), . . . , Bm (t )) is an m-dimensional Brownian motion process and 0 j (s, t )X (s)dBj (s), j = 1, 2, . . . , m, are the
It integrals.
We approximate X (t ), f (t ), (s, t ) and j (s, t ), j = 1, 2, . . . , m, by relations (9), (10) as follows

X (t ) X T (t ) = T (t )X ,
f (t ) F T (t ) = T (t )F ,
(s, t ) T (s) (t ) = T (t )T (s),
j (s, t ) T (s)j (t ) = T (t )jT (s), j = 1, 2, . . . , m.
 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143 137

In the above approximates, X and F are stochastic block pulse coefficient vectors, and and j , j = 1, 2, . . . , m, are
stochastic block pulse coefficient matrices.
With substituting above approximation in Eq. (22), we get

t m
t
X (t ) F (t ) + X
T T T
(s) (s)ds (t ) + X
T T
(s) (s)dBj (s) j (t ).
T
(23)
0 j =1 0

Let i and ji , j = 1, 2, . . . , m be the ith row of the constant matrices and j , for j = 1, 2, . . . , m; Ri be the ith row of
the integration operational matrix P; RiS be the ith row of the stochastic integration operational matrix PS ; Di and D i are
j

diagonal matrices with i and ji , j = 1, 2, . . . , m, as its diagonal entries. By the previous relations and assuming m1 = m2 ,
we have,

t t
(s) (s)ds (t ) =
T
(s) (s)ds (t )
T

0 0

R1 (t )1 (t )

R1 D1
R2 (t )2 (t ) R2 D 2

= .. = . (t ) = A (t ), (24)

. ..
R (t ) (t )
m m
Rm Dm
where
11 212 213 21m

0 22 223 22m
h 0
A= 0 33 23m , (25)
.. .. .. ..

2 ..
. . . . .
0 0 0 mm mm

and also for the It integral terms, we have

t t
(s) (s)dB(s) j (t ) =
T
(s) (s)dB(s) j (t )
T

0 0

R1S D 1

R1S (t )j1 (t )

j
R2 (t ) 2 (t ) R2 D
S j S j2
= .. = . (t ) = Aj (t ), (26)

..

.
S (t )j (t )
Rm m
Rm
S D
m
j

where
j h j j j

11 B 12 B(h) 13 B(h) 1m B(h)
2
j 3h j j

0 22 B B(h) B(2h) B(h)
23 2m B(2h) B(h)
2


j 5h j .

Aj = 0 0 33 B B(2h) 3m B(3h) B(2h) (27)
2


.. .. .. .. ..
. . . . .


(2m 1)h

0 0 0 mm
j
B B((m 1)h)
2

With substituting relations (24) and (26) in (23), we get

m

X (t ) F (t ) + X A (t ) + X
T T T T
Aj (t ).
j =1

Then,

m

X T
I A Aj FT. (28)
j =1
138 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143

m T
So, by setting M = I A j =1 Aj and replacing by =, we will have,

MX = F . (29)
Which is a linear system of equations with lower triangular coefficients matrix that gives the approximate block pulse
coefficient of the unknown stochastic processes X (t ).

5. Error analysis

In this section, we will show that the rate of convergence presented method for solving stochastic ItVolterra integral
equations with several independent white noise sources is O(h).

Theorem 1. Suppose that f (t ) is an arbitrary real bounded function, which is square integrable in the interval [0, 1), and
e(t ) = f (t ) fm (t ), t I = [0, 1), which fm (t ) = i=1 fi i (t ) is the block pulse series of f (t ). Then,
m

h
e(t ) sup |f (t )|. (30)
2 3 t I

Proof. Let,

f (t ) fi t Di ,

ei (t ) = (31)
0 t I Di

where Di = {t : (i 1)h t < ih, h = 1

m
} and i = 1, 2, . . . , m.
We have,
ih ih
1 1
ei (t ) = f (t ) f (s)ds = f (t ) f (s) ds,
h (i1)h h (i1)h
now by mean value theorem, we get,
f (i ) ih

1
ei (t ) = (t s)ds = f (i ) t + i + h , t , i Di , i = 1, 2, . . . , m
h (i1)h 2
then,
ih ih
2
2 1
ei (t )2 = |ei (t )|2 dt = f (i )

t+ i + h dt
(i1)h (i1)h 2

h3 2
= f (i ) , i Di , i = 1, 2, . . . , m. (32)
12
Consequently
1 2
1 m
e(t ) =2
|e(t )| dt = 2
ei (t ) dt
0 0 i=1
1
m

m
1 m

= e2i (t ) + 2 ei (t )ej (t ) dt = e2i (t )dt = ei (t )2
0 i=1 i <j i =1 0 i =1

m
h3 2 h2
= f (i ) sup |f (t )|2 , (33)
12 i=1 12 t I
or,
h
e(t ) sup |f (t )|
2 3 t I
hence, e(t ) = O(h). 

Theorem 2. Suppose that f (s, t ) L2 [0, 1) [0, 1) and e(s, t ) = f (s, t ) fm (s, t ), (s, t ) D = [0, 1) [0, 1), which

fm (s, t ) = i (s)j (t ) is the block pulse series of f (s, t ). Then,

m m
i=1 j=1 fij

h 21
e(s, t ) sup |fs (x, y)|2 + sup |ft (x, y)|2 . (34)
2 3 (x,y)D (x,y)D
 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143 139

Proof. Let,

f (s, t ) fij (s, t ) Dij ,

eij (s, t ) = (35)
0 (s, t ) D Dij

where Dij = {(s, t ) : (i 1)h s < ih, (j 1)h t < jh, h = 1

m
} and i, j = 1, 2, . . . , m.
For i, j = 1, 2, . . . , m, we have,
ih jh ih jh
1 1
eij (s, t ) = f (s, t ) f (x, y)dydx = f (s, t ) f (x, y) dydx,
h2 (i1)h (j1)h h2 (i1)h (j1)h

now by mean value theorem, we get,

ih jh
1
eij (s, t ) = (s x ) f s

(i , j ) + (t y) ft

(i , j ) dydx
h2 (i1)h (j1)h

1 1
= fs (i , j ) s + i +

h + ft (i , j ) t + j +

h , (s, t ), (i , j ) Dij
2 2
then,
ih jh
eij (s, t )2 = |eij (s, t )|2 dtds
(i1)h (j1)h

h4 2
= fs (i , j ) + ft2 (i , j ) , (i , j ) Dij , i, j = 1, 2, . . . , m. (36)
12
Consequently
1 2
1 1 1 m
m
e(s, t ) =
2
|e(s, t )| dtds = 2
eij (s, t ) dtds
0 0 0 0 i =1 j =1
m
m 1 1 m
m
= e2ij (s, t )dtds = eij (s, t )2
i=1 j=1 0 0 i=1 j=1

m m
h4 2 h2
= fs (i , j ) + ft2 (i , j ) sup |fs (x, y)|2 + sup |ft (x, y)|2 , (37)
12 i=1 j=1 12 (x,y)D (x,y)D

or,

h 21
e(s, t ) sup |fs (x, y)|2 + sup |ft (x, y)|2
2 3 (x,y)D (x,y)D

hence, e(s, t ) = O(h). 

Theorem 3. Suppose X (t ) is the exact solution of (22) and Xm (t ) is the block pulse series approximate solution of (22) that their
coefficients are obtained by (29). Also assume that
(1) X (t ) , t I = [0, 1),
(2) (s, t ) M , (s, t ) D = I I,
(3) j (s, t ) Mj , j = 1, 2, . . . , m (s, t ) D = I I,
m
(4) M + (h) + j =1 Mj + j (h) supt I |Bj (t )| < 1,

then
m
(h) + (h) + j (h) sup |Bj (t )|

j =1 t I
X (t ) Xm (t ) , t I (38)
m
M + (h) + Mj + j (h) sup |Bj (t )|

1
j =1 t I

where
12
X (t ) = E X 2 (t ) ,
140 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143

h
(h) = sup |f (t )|,
2 3 t I
h 21
( h) = sup |s (x, y)|2 + sup |t (x, y)|2 ,
2 3 (x,y)D (x,y)D

h 12
j (h) = sup |js (x, y)|2 + sup |jt (x, y)|2 , j = 1 , 2 , . . . , m.
2 3 (x,y)D (x,y)D

Proof. By using Theorems 1 and 2, we have,

h
f (t ) fm (t ) sup |f (t )| = (h), (39)
2 3 t I
and

h 21
(s, t ) m (s, t ) sup |s (x, y)|2 + sup |t (x, y)|2 = (h), (40)
2 3 (x,y)D (x,y)D

and
h 21
j (s, t ) jm (s, t ) sup |js (x, y)|2 + sup |jt (x, y)|2 = j (h), j = 1, 2, . . . , m. (41)
2 3 (x,y)D (x,y)D

From (22), we get

t
X (t ) Xm (t ) = f (t ) fm (t ) + (s, t )X (s) m (s, t )Xm (s) ds
0
m t

+ j (s, t )X (s) jm (s, t )Xm (s) dBj (s), (42)
j =1 0

then by mean value theorem, we can write

X (t ) Xm (t ) f (t ) fm (t ) + t (s, t )X (s) m (s, t )Xm (s)

m
+ Bj (t )j (s, t )X (s) jm (s, t )Xm (s). (43)
j=1

By using (H1), (H2) and (40), we have

(s, t )X (s) m (s, t )Xm (s) (s, t ) X (s) Xm (s) + (s, t ) m (s, t ) X (s) Xm (s) + X (s)

(M + (h))X (s) Xm (s) + (h), (44)

and
j (s, t )X (s) jm (s, t )Xm (s)

j (s, t ) X (s) Xm (s) + j (s, t ) jm (s, t ) X (s) Xm (s) + X (s)

(Mj + j (h))X (s) Xm (s) + j (h), j = 1, 2, . . . , m. (45)

So

X (t ) Xm (t ) (h) + t M + (h) X (t ) Xm (t ) + (h)
m

+ Bj (t ) Mj + j (h) X (t ) Xm (t ) + j (h) , (46)
j=1

or
m
(h) + (h) + j (h) sup |Bj (t )|

j =1 t I
X (t ) Xm (t ) , t I (47)
m
M + ( h) + Mj + j (h) sup |Bj (t )|

1
j =1 t I

hence, X (t ) Xm (t ) = O(h). 
 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143 141

Table 1
Mean, standard deviation and mean confidence interval for error in Example 1 with n = 20.
m xE sE 0.95 confidence interval for mean of E
Lower Upper

4 2.80383739E4 1.42692696E4 2.17845908E4 3.42921571E4

8 4.89201948E4 1.42851647E4 4.26594454E4 5.51809443E4
16 6.52973571E4 1.10050413E4 6.04741850E4 7.01205291E4
32 0.00121313404 2.14221469E4 0.00111924509 0.00130701849
48 0.00172837457 1.92647371E4 0.00164394314 0.00181280600
64 0.00218103587 2.38125316E4 0.00207667285 0.00228539889
80 0.00254096373 3.96350688E4 0.00236725539 0.00271467208

Table 2
Mean, standard deviation and mean confidence interval for error in Example 1 with n = 20.
m xE sE 0.95 confidence interval for mean of E
Lower Upper

4 2.80383739E4 1.42692696E4 2.17845908E4 3.42921571E4

8 4.89201948E4 1.42851647E4 4.26594454E4 5.51809443E4
16 6.52973571E4 1.10050413E4 6.04741850E4 7.01205291E4
32 0.00121313404 2.14221469E4 0.00111924509 0.00130701849
48 0.00172837457 1.92647371E4 0.00164394314 0.00181280600
64 0.00218103587 2.38125316E4 0.00207667285 0.00228539889
80 0.00254096373 3.96350688E4 0.00236725539 0.00271467208

6. Numerical examples

To illustrate the method stated in Section 4, we consider below some examples. The computations associated with the
examples were performed using Matlab 7. Let Xi denote the block pulse coefficient of exact solution of the given examples,
and let Yi be the block pulse coefficient of computed solutions by the presented method. The error is defined as
E = max |Xi Yi |.
1im

Example 1. Consider the following linear stochastic ItVolterra integral equation,

t m
t
X (t ) = X0 + rX (s)ds + j X (s)dBj (s) s, t [0, 1), (48)
0 j =1 0

m m
(r 1 2 )t + B (t )
with the exact solution X (t ) = X0 e 2 j=1 j j=1 j j , for 0 t < 1, where X (t ) is an unknown stochastic processes
defined on the probability space ( , z, P ), and B(t ) = (B1 (t ), B2 (t ), . . . , Bm (t )) is an m-dimensional Brownian motion
process. The numerical results for X0 = 200 1
, r = 201
, 1 = 50 1
, 2 = 50 2
, 3 = 50 4
and 4 = 50
9
are shown in Table 1.
In Table 1, n is the number of iterations, xE is the mean of error, and sE is the standard deviation of error. The curves in
Fig. 1 represent a trajectory of the approximate solution computed by the presented method with a trajectory of the exact
solution. The curves in Fig. 2 represent the variation process of error.

Example 2. Consider the following linear stochastic ItVolterra integral equation,

t m
t
X (t ) = X0 + r (s)X (s)ds + j (s)X (s)dBj (s) s, t [0, 1), (49)
0 j =1 0

t m m t
(r (s) 1 2 (s))ds+ (s)dB (s)
with the exact solution X (t ) = X0 e 0 2 j=1 j j=1 0 j j
, for 0 t < 1, where X (t ) is an unknown stochastic
process defined on the probability space ( , z, P ), and B(t ) = (B1 (t ), B2 (t ), . . . , Bm (t )) is an m-dimensional Brownian
motion process. The numerical results for X0 = 12 1
, r (s) = s2 , 1 (s) = sin(s), 2 (s) = cos(s) and 3 (s) = s are shown in
Table 2. In Table 2, n is the number of iterations, xE is mean of error, and sE is the standard deviation of error. The curves
in Fig. 3 represent a trajectory of the approximate solution computed by the presented method with a trajectory of exact
solution. The curves in Fig. 4 represent the variation process of error.

7. Conclusion

Because it is almost impossible to find the exact solution of Eq. (22), it would be convenient to determine its numerical
solution based on stochastic numerical analysis. Using block pulse functions as basis functions to solve the linear stochastic
142 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143

Fig. 1. The trajectory of the approximate solution and exact solution of Example 1 for m = 32, m = 80, n = 20.

Fig. 2. Variation trend of error in Example 1 for m = 48, n = 20, n = 50.

Fig. 3. The trajectory of the approximate solution and exact solution of Example 1 for m = 32, m = 80, n = 20.

ItVolterra integral equations with several independent white noise sources is very simple and effective in comparison
with other methods. Its applicability and accuracy is checked on some examples.
The advantages of using the block pulse functions are simple calculations and conversion of the integral equation to a
triangle system. So, by using block pulse functions and their stochastic operational matrix for stochastic ItVolterra integral
equation, oscillations appear. Fluctuations generated in the exact and approximated answers are not dependant on a selected
basis, but on random factors in the equation. Oscillations that appeared in the answers are due to It integrals in the integral
equation. It integral is related to the Brownian motion. Simulation of Brownian motion B(t ), is related to random numbers
with normal distribution, since random number have fluctuations, these fluctuations are transmitted to the responses. With
this method, different random paths for simulated Brownian motion is obtained, and we therefore consider the average of
the directions. We used the article [17] for the simulation of Brownian motion.
 K. Maleknejad et al. / Computers and Mathematics with Applications 63 (2012) 133143 143

Fig. 4. Variation trend of error in Example 1 for m = 48, n = 20, n = 50.

Moreover, one could also apply the ItTaylor expansion described by Kloeden and Platen [3], or those from article [18],
for example. Certainly, it could be the topic of some future work.

References

[1] M. Khodabin, K. Maleknejad, M. Rostami, M. Nouri, Numerical solution of stochastic differential equations by second order RungeKutta methods,
Mathematical and Computer Modelling 53 (2011) 19101920.
[2] M. Khodabin, K. Maleknejad, M. Rostami, M. Nouri, Interpolation solution in generalized stochastic exponential population growth model, Applied
Mathematical Modelling, in press, Corrected Proof, Available online 23 July 2011.
[3] P.E. Kloeden, E. Platen, Numerical Solution of Stochastic Differential Equations, in: Applications of Mathematics, Springer-Verlag, Berlin, 1999.
[4] J.C. Cortes, L. Jodar, L. Villafuerte, Numerical solution of random differential equations: a mean square approach, Mathematical and Computer
Modelling 45 (2007) 757765.
[5] B. Oksendal, Stochastic Differential Equations: An Introduction with Applications, fifth ed., Springer-Verlag, New York, 1998.
[6] K. Maleknejad, M. Khodabin, M. Rostami, Numerical solution of stochastic Volterra integral equations by a stochastic operational matrix based on
block pulse functions, Mathematical and Computer Modelling, in press, Corrected Proof, Available online 14 September 2011.
[7] M.A. Berger, V.J. Mizel, Volterra equations with Ito integrals I, Journal of Integral Equations 2 (1980) 187245.
[8] X. Zhang, Euler schemes and large deviations for stochastic Volterra equations with singular kernels, Journal of Differential Equations 244 (2008)
22262250.
[9] S. Jankovic, D. Ilic, One linear analytic approximation for stochastic integro-differential equations, Acta Mathematica Scientia 30 (2010) 10731085.
[10] X. Zhang, Stochastic Volterra equations in Banach spaces and stochastic partial differential equation, Journal of Functional Analysis 258 (2010)
13611425.
[11] K. Maleknejad, H. Safdari, M. Nouri, Numerical solution of an integral equations system of the first kind by using an operational matrix with block
pulse functions, International Journal of Systems Science 42 (2011) 195199.
[12] K. Maleknejad, Y. Mahmoudi, Numerical solution of linear Fredholm integral equation by using hybrid Taylor and block-Pulse functions, Applied
Mathematics and Computation 149 (2004) 799806.
[13] K. Maleknejad, S. Sohrabi, Y. Rostami, Numerical solution of nonlinear Volterra integral equations of the second kind by using Chebyshev polynomials,
Applied Mathematics and Computation 188 (2007) 123128.
[14] K. Maleknejad, B. Rahimi, Modification of block pulse functions and their application to solve numerically Volterra integral equation of the first kind,
Communications in Nonlinear Science and Numerical Simulation 16 (2011) 24692477.
[15] Z.H. Jiang, W. Schaufelberger, Block Pulse Functions and Their Applications in Control Systems, Springer-Verlag, 1992.
[16] G. Prasada Rao, Piecewise Constant Orthogonal Functions and their Application to Systems and Control, Springer, Berlin, 1983.
[17] Desmond J. Higham, An algorithmic introduction to numerical simulation of stochastic differential equations, in: Society for Industrial and Applied
Mathematics, SIAM Review 43 (3) (2001) 525546.
[18] C. Tudor, M. Tudor, Approximation schemes for ItoVolterra stochastic equations, Boletin Sociedad Matemtica Mexicana 3 (1) (1995) 7385.