Keywords: rough surface scattering; binary space partitioning; ray tracing; nearest object priority; communications. I. INTRODUCTION. Proper modeling of indoor ...
STRATEGIES AND SCHEMES
Nearest Object Priority Based Integrated Rough Surface Scattering Algorithm for 3D Indoor Propagation Prediction ASMZ Kausar, AW Reza, KA Noordin, MJ Islam,
H
Ramiah
Department of Electrical Engineering, Faculty of Engineering, University of Malaya, Kuala Lumpur 50603, Malaysia
Abstract:
Since rough surface scattering
and ray prediction accuracy are also presented.
has a great impact on the accuracy of
Keywords:
rough surface scattering; binary
the propagation prediction algorithm, an
space partitioning; ray tracing; nearest object
integrated algorithm for indoor propagation
priority; communications
prediction including rough surface scattering is proposed here. This algorithm is composed
I. INTRODUCTION
of a three dimensional (3�) ray tracing algorithm based on binary space partitioning
147
Proper modeling of indoor wireless commu
(BSP) and a diffuse scattering algorithm based
nication system is necessary for achieving a
on Oren-Nayar's theory. Lack of accuracy and
higher degree of service quality. Rough sur
prohibitive time consumption are the main
face scattering is one of the significant propa
drawbacks of the existing ray tracing based
gation mechanisms,which has the capability
propagation prediction models. To defy these
to change the amount of the received signal by
shortcomings,the balanced BSP tree is used
changing the ray directions. For that reason,
in the proposed algorithm to accelerate the
rough surface scattering has a great impact on
ray tracing,while the nearest object priority
indoor propagation planning. Besides,proper
technique (NOP) and in contact surface
understanding about the scattering mechanism
(lCS) is used to eliminate the repeated ray
has a great role in target identification,remote
object intersection tests. Therefore,the final
sensing, radio astronomy, and radar design.
criteria of this study are the time consumption
Consequently, accurate and efficient rough
as well as accuracy by predicting the field
surface scattering analysis is an important re
strength and the number of received signals.
search area in electromagnetic.
Using the proposed approaches,our algorithm
Because of the multi-interaction between
becomes faster and more accurate than the
rays and object, scattering analysis is very
existing algorithms. A detailed comparative
complicated and difficult. A lot of research
study with existing algorithms shows that
work is progressing for the solution of rough
the proposed algorithm has at most 37.83%
surface scattering [1-14]. Among those,some
higher accuracy and 34.44% lower time
are experimental model [5,7],some are statis
consumption. Moreover,effects of NOP and
tical model [4,9],and some are deterministic
ICS techniques and scattering factor on time
model [3,8] where deterministic models are China Communications· October 2014
popular for indoor propagation modeling.
is very time consuming because of the trun
Therefore, less prediction time and higher
cated rough surface in KA-MoM. This surface
In this paper, an inte
accuracy are prime concerns for determinis
is usually very large in terms of wavelength,
grated technique is
tic model. However,the existing techniques,
particularly for low incident angles. In near
such as shooting and bouncing ray (SBR)
field technique,the edge diffraction is not con
[15],bi-directional path tracing (BDPT) [16],
sidered. However,this causes lower accuracy.
brick tracing (BT) [17],radiosity [18],Kirch
Hence, to improve the performance of the
hoff Approach-Method of Moments (KA
ray tracing algorithm,researchers use some ac
MoM) [19],and near field [14] require higher
celeration techniques. Most acceleration tech
prediction time due to complex algorithms
niques have been used in two different ways
used. Moreover, the prediction accuracy is
[14]. To simplify the simulation environment
not so high. In SBR technique,the spherical
is one of the techniques to achieve accelera
wavefront is covered by the ray cones. Two
tion,while intersection test calculation is the
successive cones overlap each other at the
other possible technique. For simplification of
edge. In that case,the receiver (Rx) located
the simulation environment,the edges,planes
in the overlapping area will then receive two
or whole objects are considered to reduce. For
rays and double ray counting error will be oc
achieving acceleration,sometimes the feature
curred. The BDPT technique shows incorrect
of three-dimensionality is neglected or the
results for single floor multiple room environ
number of objects involved is reduced. The
ments,where more than one transmitter ( Tx) is
intersection test calculation is needed to find
present. In that case,it also takes a lot of time
out the actual rays. For this,all possible paths
to create the ray paths. In BT technique,every
between Tx and Rx are tested to find the actual
wall of the building is discretized into its unit
path, and then ray tracing is conceded. This
cells,called bricks. Because of the truncation
path finding test is a matter of acceleration and
of the slab,the corner bricks show factual er
a numerous researches have been conducted
ror,which results in erroneous analytic reflec
for this purpose,such as the efficient visibili
tion and transmission coefficient. Furthermore,
ty graphs both in 2D [20] and 3D cases [21],
BT technique only considers the walls of the
hybrid approaches with a 2D ray-path search
indoor environment. Without considering the
together with 3D ray tracing [22],etc.
objects,the accurate results cannot be found.
In this paper, a 3D ray tracing technique
Besides, in BT technique,after the first in
integrated with a scattering technique is pro
teraction between Tx and bricks,each related
posed,considering each interaction point of
bricks become new source of the ray. In that
a ray. The binary space partitioning (BSP)
case,all of those bricks have to consider as Tx
based technique is used for ray tracing,which
individually and test for the next interaction.
consumes less time than conventional ray
This process consumes a huge amount of time.
tracing technique. Oren-Nayar [23] scattering
The radiosity technique cannot be used for
technique is used to determine the reflected ra
smooth surfaces. Hence,for the objects having
diance of the scattered rays and backward-for
a smooth surface,this technique offers inaccu
ward technique [24] is used for the simulation
rate results. Additionally, radiosity technique
purpose. These techniques are chosen for their
is mainly used for acoustic waves, not for
simplicity,accuracy,and ease of usability. The
electromagnetic waves. The propagation prop
type of the ray-object contact point and dis
erties of different media except air (e.g.,walls,
tinguished object will make a variation to the
woods, bricks, concretes) will be of course
scattering parameters. Therefore,we have also
different for either type of wave and cannot be
considered different types of objects. For min
interchanged,which results in lack of accura
imizing the intersection test time,a new tech
cy. In KA-MoM technique, the coupled field
nique,namely,nearest object priority (NOP) is
calculation between object and rough surface
introduced here. In the existing algorithms,all
China Communications· October 2014
presented to develop a new indoor prop agation prediction algorithm including rough surface scatter ing.
148
of the objects should be tested for finding the
surfaces, which overcomes the limitation of
actual object. Therefore,the prediction time
radiosity technique. Diffraction from the edge
becomes extremely high. In our algorithm,we
of an object is also considered in the proposed
have a NOP technique for finding the nearest
technique which was not considered in KA
object, which will minimize the intersection
MoM technique.
testing time. For reducing 3D computational
The mathematical approach of the proposed
complexity and saving the prediction time,we
hybrid technique along with the NOP and ICS
have introduced the novel concept of in con
are presented in Section 2,while detailed re
tact surface (lCS).
sults analysis are depicted in Section 3. Final
As stated above,the major problems of ray
ly,a conclusion is made in Section 4.
tracing are ray-object intersection test time and accuracy. Therefore,the main contributions of
II. PROPOSED HVBRIDE TECHNIQUE
this paper are(i) The NOP and ICS for reducing the ray-object intersection test time. (ii) The enhancement of accuracy by intro ducing the ICS and rough surface scattering.
2.1 Scattering technique
The proposed hybrid technique is composed of scattering and ray tracing techniques. As we
(iii) The elimination of the limitations of
know that the scattering model cannot be used
the existing techniques (as described above).
directly with a ray tracing model [18],there
The double ray counting problem of SBR tech
fore,we have used the backward and forward
nique is overcome by using single ray instead
scattering function of objects and walls. On
of ray cone. The problem of incorrect results
the other hand,the Oren-Nayar [23] roughness
due to multiple rooms in BDPT technique is
technique is used to find out the reflected radi
solved by using the proper use of NOP and
ance,which has a usage to find out the direc
ICS and thus,the proposed technique shows
tion of the scattered rays.
decent results not only for single rooms but
First of all,the target is to find out the re
also for multiple room environments. Unlike
flected radiance from the roughness technique,
BT technique,we have considered walls and
proposed by Oren-Nayar [23,25],which is the
objects of the environment for reducing the
modification of the Lambertian method [25,
chance of missing the rays due to the different
26]. Figure 1 shows different elements used to
ray-object interactions. Our proposed tech
find out the reflected radiance. The reflected
nique can be used for both smooth and rough
radiance can be changed due to the presence of masking, shadowing,and inter-reflection effects. In case of shadowing,a facet is just partly illuminated due to the shadow of neigh
z
boring facet. As the nearby facets create an
v
obstruction,the sensor can detect a facet partly 6r
because of masking. In inter-reflection,rays 6i
s
bounce between adjacent facets. In Figure 1, 8i represents the incidence =---��
r
angle and CPi represents an incident azimuth angle,while 8r denotes the reflected angle and CPr denotes the reflected azimuth angle. The
terminology of diffuse reflection is expressed in terms of reflected radiance Lr and incident radiance Li• Now,let us consider an isotropic surface of Fig.1 The geometry used to characterize the reflected radiance
149
V
cavities with same facet slop 8a
and uniform distribution in orientation CPa. The China Communications· October 2014
isotropic surface of same facet slop is chosen
to determine the direction of the scattered
instead of different facet slop on behalf of
rays. Since the surface is considered as isotro
reducing the computational complexities. Ac
pic,the value of reflected radiance and other
cordingly,the radiance is determined as [23,
parameters except the incident and reflected
25]-
angle will be remain unchanged in all inci
L: (B",Bj, cP, - CPj;tT)
dences if the object type is not changed. Now, the reflected angle is found by simply putting
K, (tT)+ cos (cp, - cP')· P
I
K2 (a;f3;cp, - cpj;tT)tanf3
--; Eo cos Bj +(1 - Icos (cp, - cP,) )· l
=
K3(a;f3;tT)tan
l,
(I)
(a+f3 ) -
Now,for prediction modeling,a scattering environment. This scattering cluster is used to find the backward and forward function
=
K,
=
K3
0.45
f 1
=
0.45
cT
(
"::� "::�
.09
(T'
.09
2
sin a if cos(rp, - rp,)
;;;. 0
(sin ( 2! n otherwise' ) ( 4af3 )2 9 ---;[2 ' a -
0.125 2 tT tT +0.0
Here,(J is the standard deviation as a mea sure of surface roughness, a
13
=
Min [Bj,B,], and
p
the scattered ray will be originated. cluster is used for the evaluation of the indoor
2
where the coefficients are tT2 KI 1 - 0.5 2 3 , tT +0. "
these values in Eq. (4). Thus,the direction of
=
Max [Bj,B,],
is the diffuse percentage
of walls and objects. The scattering cluster parameter varies according to the nature of interaction between the ray and the object,and the interface object type. During simulation, the simulated response is analyzed by impos ing the backward-forward function into the described ray tracing technique. The scattered direction is [24]-
for Lambert's Law. Eq. (1) is for masking and shadowing con ditions. For inter-reflection, the expression is
1
directive pattern for backward and forward
IE,12
=
[23]V, (B"B"cp, - CPj;tT) -
tT2
_
. -
2
=
p 0.17-Eo cos B, 1f
r I - cos (cp, - cP,) ( ----;213 )'1 . J I
where (2)
sulting from all effects is [23], =
•
"
EsoT'
IJfR
cos 1/1" 2 + cos I/Ir 2
+
and
IJfT
)
'"
)
"
' For backward direction
"'
; For forward direction '
(5) are the angle between the
scattering direction and the reflected or trans
mitted ray direction,respectively,aR and aT are related to scattering lobe,and
Therefore,the total reflection radiance re L,(B"B"cp, - CPj;tT) L�(B"B"cp,CPj;tT)+L,2 (B"Bj,cp, - CPj;tT)
(1 2 (1
EsoR'
(3)
A modification has been done based on the
ESOR
and
ESOT
is
dependent on SR and S),respectively. The pre eminent values of SR and
aR
for building walls
are 0.4 and 3,respectively and the best values for S) and
a)
for a brick-wall sample are 0.5
and 3,respectively [24,27].
term K3 by assuming that there is no inter-re
2.2 Ray tracing with proposed NOP
flection and the modified equation is as [23],
and ICS techniques
[A�,Max I+
L, (B"Bj, cP, - CPj;tT)
=
� Eo cos Bj
Al
where,
A2
=
=
1
[0, cos (B, - B,)]. '
smatanf3
In conjunction with the above backward and (4)
tT2
1 - 0.5� 33' . tT2 0.45 2 tT +0.09 .
The simplified Eq. (4) has the advantages
forward function, a ray tracing technique is used to predict the propagation path. For the proposed ray tracing algorithm,we have used the balanced BSP technique to minimize the simulation time and also to increase the ac curacy. In this technique, the objects of the indoor environment are arranged in a tree on the basis of partitioning until the smallest unit is found. Our partitioning algorithm splits the
China Communications· October 2014
150
subdivision by choosing the planes which are normal to any of the 3 axes in a round robin The BSP is a technique that recursively splits the target space using random plane. Mathematically [28][1,2],we can say that,if S be a group of n pair-wise separate objects in 3D space IR3, a binary space division parti tion for S is a recursively demarcated convex subdivision of space acquired by dividing the space into two uncluttered regions SI and S2 by a plane. Thus, it recursively constructs a
BSP for {s n S lis E S} within SI and a BSP n
and let Pi be a random plane orthogonal to the
j-th coordinate path. Let Pi and Pi indicates
technique.
for {s
jects in a 3D space 1R3 shown in Figure 2(b),
S21s E S} within S2. This process
comes to an end when each cell of the BSP tree splits up at most single object of S. For constructing a balanced tree,we have used the window list concept. This window list concept
the set of objects lying in the part below the cut Pi and above the cut Pi' respectively. Next,
let Pi- indicates the set of objects,which are
intersected by the plane Pi' Let,the maximum
number of objects intersected by the plane Pi
kj . The profile k* of S is demarcated to be the minimum of all k/, i.e.,
is
'
k'
object-splitter intersection or not. An axis par [ 29 ]:
R
=
[a),b)]
X
[a2,bJ
X
[a3,b3].
(6)
Any plane P normal to any of the 3 axes
divides IR3 into two halves as shown in Fig ure 2(a). Any object R intersected by plane P
splits into two parts,which are non-overlap ping. More specifically,for any P E (ai' b), the plane xi=P splits the object R into two parts:
R"It
=
Right
=
[a),b)]
X
[a),b)]
x
[a2,bel [a2,bel
x
x
[a3,b3]
•••
[a3,b3]
•••
x
x
[ai,p], (7) [p,bJ(8)
Now,let S ={R],R2, ... ,Rn}={[a],b]],[a2,b2], ... , [a",bnD be a set of n feasibly overlapping ob-
min{k;, k;, k;}.
(9)
Let,l=mini at. r=maxi bi for 1:S i:S n , and x= be the set of objects in S close to x for each x E [1, r].x· is named the profile of S at x. Now, [I, r] will split into a non-overlapping set of T in parts of/],... ,m I and it is satisfied by the fol
lowing [29 ]:
I Wi l � ok' (Wi),
will serve for all the cases whether there is an allel obstacle R in a 3D space 1R3 is defined as
=
(10)
w h e r e 62:2 i s a c o n s t a n t a n d k'(W) max(l,minl x· I) i s t he min im u m tE/i =
profile of the set Wi' Each set Wi' for 1:S j
:s
m, is termed as window and the systematic sequence W=( W ],W b "" Wn,) is called the win
dow list W(IfI). Any inquiry series q=[q.Xl ,q.
x2]x[q.y],q'Y2] can be disintegrated into a max imum of three definite parts q" qm, and qn"The
final algorithm to create the balanced BSP tree is shown in Algorithm I . We can describe the proposed BSP algo rithm (Algorithm 1) in a simple and straight forward approach as below(i) First,select a partition plane, (ii) Partition all objects with the initial par tition plane,storing them in either the front list
s
or back list,and D
y
z
D
p
I�I�·I (a)
D x
(iii) Iterate through the front list and back
O
list,creating a new tree node,and attaching it to the left or the right leaf of the parent node.
�
There are four cases to deal with when an environment is partitioned by the plane:
o
(i) Object is in front of the plane, (ii) Object is behind the plane, (iii) Object is coincident with the plane,
F' j �
(iv) Object spans the plane. (b)
Fig. 2 Illustration of creation of window list (a) Single object (b) Multiple object
15 1
For the first two cases,the object is simply added to the appropriate node of the tree. For China Communications· October 2014
the last two cases,window list will be applied for adding the object in the back list or front list. The ICS (klmn),shown in Figure 3(a) is de fined as the effective surface within an object. The 3D objects we used in this study are made of cubes or cuboids as in Figure 3,which con sist of six faces and eight vertices,where each vertex has a unique coordinate point. Using these vertices of the object,we can calculate abscissa of Cl abscissa of C3 ,(\\) =
ordinate of k
----
=
=
ordinate of Cl ordinate of C3"
(12 )
I, m,and n coordinates can also be determined
in the same manner. Among all of the objects in an environ ment,only the closest object intersects with a specific ray. The NOP technique helps us to find out this closest object. To fonnulate NOP in Figure 3(b),we assume a vector q started from the ICS to the origin tween
0
0,
distance s be
and plane P, vector I, which has a
unity length,and an angle Q between q and I.
From the vector analysis [30],we know that, a dot product is equal to the product of the ab solute length of the two vectors,multiplied by the angle between them or sum of the product of the components. Mathematically,it can be
=
=
Iql cos Q
Input: a set of objects Output: a BSP tree of the objects 1. {plane partition;
2. list polygons; 3. BSP_tree
4.
q.x * l.x + q.y * l.y
+
. (\3) q.z * l.z
*front, *back;};
5. void Construct_BSP_Tree(BSP_tree *tree, list polygons)
6. { polygon *root=polygon.GetJrom_List 0;
8.
9.
I O.
tree->partition = root->Get]lane 0; tree->polygons.Add_To_List(root); list front_list, back_list;
11. polygon *poly;
12. while «poly=polygons.GetJrom_List()) !=O)
13. { int result=tree->partition.Classify_Polygon (poly); 14.
15. { 16.
switch(result) case IN_BACK_OF: backlist.Add_To_List(poly);
17.
break:
18.
case IN FRONT OF:
19.
-
-
frontlist.Add_To_List(poly);
W.
break:
21.
case SPANNING:
n.
polygon *front�iece, *back�iece;
�. �.
SplicPolygon(poly, tree->partition, front_piece, back_piece);
�.
�.
frontlist.Add_To_List(front_piece);
backlist.Add_To_List(back�iece); break})
n.
28. { �. 30.
if(! fronUist.Ts_EmptLList()) tree->front=new BSP tree: ConstrucCBSP_Tree(tree->front, fronUist);}
31. {
expressed as follows: dot (q, l)
Construct BSP Tree(Polygon Set) Assumption: a specified technique for building binary tree and a root plane
7.
the coordinate points of k as follows:
abscissa of k
Algorithm 1 The proposed balanced BSP algorithm
if(! backUist.ls_Empty_List()) tree->back=new BSP_tree; Construct_BSP_Tree(tree->back, back_list);})
However,one of the elements of vector q, IqlcosQ is in the direction of vector I. That
the object type can also be identified. Based on
means,if dot(q, l) is a less significant quan
this identification, decision of the occurrence
tity than s,top of the plane is denoted and if
of reflection, refraction,and diffraction will
dot(q, l) is superior in respect to s,bottom of
be taken and the next ray will be generated ac
the plane is denoted. Therefore,by subtracting
cordingly.
them,we found: Side
Side
=
=
sgn (s
Referring to Figure 3(b),suppose Ll is the - dot (q, I)),
(\4)
sgn(s-(q.x*l.x + q.y*l.y + q.z*l.z)). (\5)
source ray and L2 is the ICS (klmn) generated after the source ray incident to the 3D object. Here,(xj,Yj,Zj) with vector (aj,bj,cj) represents
of any object is facing the Tx by using Eq.
Ll, while (X2,Y2,Z2) with vector (a2,b2,c2) rep resents L2 or the [CS. Now,if these two lines
(15). If we can identity the nearest object then
intersect with each other,then the intersection
Now,we can quickly determine which side
China Communications· October 2014
152
to Rx. In this figure,only two rays Ll and L2 are shown as significant in the environment where multipath propagation takes place. III. RESULTS AND DISCUSSION
In the following,the simulation results of 3D (a)
indoor environments with rough surface scat
Intersection point between Ll and ICS p Cl
tering are presented. The simulation results will be changed according to the scattering
__
factor [31],which varies for different objects. Therefore, during simulation, different types of objects have been used. The comparison is made between the proposed technique and the existing techniques, such as radiosity [18],near-field [14],KA-MoM [19],ST [17], BDPT [16], and SBR [15] techniques. For this comparison,we have used five different
(b)
indoor environments. The details of five dif Fig. 3 (a) Creation of lCS (b) Fundamentals of NOP
ferent realistic indoor environments are shown in Figure 4. The experimental settings of all algorithms are kept analogous to make a fair
point can be calculated. By using the cross
comparison. Some important parameters are
product [30] of these two vectors,we can find
given in Table I. The simulation is carried out for 10 different scenarios for each of the five
whether these lines are intersecting or not.
Ll where e (0°
X
:s
L2 e
:s
=
1£111L21sin8.J],
(16)
changing the positions of Tx inside a particu
180°) is the measure of the
lar environment. For every scenario,we have
smaller angle between Ll and L2, IL11 and IL21 are the magnitudes of vectors Ll and L2, and J] is a unit vector perpendicular to the plane
time and the number of predicted signals for
collected the necessary data for the execution all of the above mentioned techniques. The
containing L1 and L2. According to vector algebra,if Eq. (16) is
are represented graphically in Figures 5 and 6,
not zero then the two lines are considered as
respectively.
results obtained from 10 scenarios of Figure 4
intersecting and the intersecting point will be
Figure 5 represents the comparison between
determined. Hence,we can write the paramet
the algorithms in terms of accuracy. The accu
ric equations for Ll and L2 as below:
racy depends on the number of rays received
L1:
x=x,+a,*tby=y,+b,*t"z=z,+c,*tj, (17)
L2:
x=x2+a2*t],y=Y2+b2*t2,Z=Z2+C] *t], (18)
where tl and tz are two unknown parameters. Now,by solving Eqs. (17) and (18), we can determine the intersection point (x ,y,z ) of L1 and L2. Now,from this point,the next ray will
by the receiver. From this point of view,the proposed algorithm shows a better accuracy than other listed algorithms and this higher accuracy is due to implementation of ICS and NOP in the proposed algorithm. In case of execution time,the proposed al gorithm shows lower time consumption as in
be generated according to the object type. In
Figure 6. Here,SSP tree is used for ray trac
an indoor environment,such as the one shown
ing,which decreases the time by arranging the
in Figure 4(a),the target here is to predict the path of the significant rays emanated from Tx 153
environments. These scenarios are created by
object details in an organized approach. The NOP technique also minimizes a huge amount China Communications· October 2014
of time for intersection test by choosing the exact object. Thus,the overall execution time becomes lower than the existing techniques. The results obtained for all five different environments are represented in Table [I. From the results, we can observe that,the proposed algorithm shows 37.83% better accuracy than SBR,20.54% better than radi
(a)
osity method, 25.40% better than near field
(b)
technique, 35.13% better than KA-MoM technique, 13.51% better than BT method, and 11.35% better than BDPT technique. It also shows 12.89% lower time consumption than SBR algorithm, 24.32% lower than ra diosity algorithm, 34.44% lower than near field technique,26.86% lower than KA-MoM
o
technique,6.67% lower than BT method,and
(e)
27.94% lower than BDPT technique. 3.1 Effects of NOP and ICS
In our projected technique,rough surface scat
rnJ ffil rID on
•
(d)
DO DD
an an
tering is included and its effects are described by using Scattering Factor (SF). SF is a sig nificant feature,which is affected by different wall materials and influences on the scattering angle. The scattering angle increases with the increase of SF. Consequently,the ray-object interaction probability increases, which raise the number of predicted signals and also the prediction time. Figure 7 shows
(e)
Fig. 4 Details a/five different indoor environments (a) Multipath propagation using
res and NOP onjirst environment (b)-(e) Remaining environments
the effect of SF on the number of predicted signals and time. The number of predicted
reducing time by managing the data efficiently
signals increases 39.94% for the increase of
than a single list. Again,NOP is reducing time
SF from 4 to 8 and 23.69% from SF 8 to 20.
by finding the nearest intersection point within
At the same time, 24.29% increased time
the shortest possible time. Thus,in two steps,
is needed for the increase of SF from 4 to 8
the time is reducing. On the other hand,for the
and 14.51% increased time from SF 8 to 20.
number of predicted signals,15.82% higher
Therefore,we can say that scattering has great
signal is predicted when NOP is included with
effect on the number of predicted signals and
scattering and 32.39% higher for inclusion of
prediction time of the ray tracing technique.
[CS. Here,NOP increases the signal predic
From Figure 8,we can easily describe that
tion by detecting the nearest object accurately
the proposed technique performs better than
and thus,it does not give any ray to be unused.
the existing techniques. [n Figure 8,step-by
Conversely, [CS increases the signal predic
step improvement is described. As of Figure
tion by handling the surface rays accurately.
8,we found that the algorithm needs 18.33% less time when BSP is included with scattering and 34.29% less time when NOP is included with both BSP and scattering. Here, BSP is China Communications· October 2014
3.2 Performance with different types of antenna
Type of antenna is an important parame154
Table I Important parameters used in experiments
for data collection. In the figures,all of the
Factor
Value
4,8,20
Scattering Factor
10 cm
Wall Thiclmess
250 cm
Wall Height
2.4 GHz
Carrier Frequency
2.15 dBi
Antenna Gain for Dipole Antenna
18 dBi
Antenna Gain for Parabolic Antenna
5.2
Permittivity of Brick
ment achieved. However, a very negligible difference has been observed,which is occur ring due to multipath propagation phenomena. This identical result with the well-established existing techniques proves that,the proposed field strength. Hence,we can conclude that the
3
Permittivity of Wood
proposed technique can perform ray tracing
4
Permittivity of Plastic (PVC Board)
strength at the receiver end and good agree
technique is accurately predicting the electric
3
Permittivity of Glass
techniques demonstrate almost similar field
accurately with lower time consumption than the existing techniques. IV. CONCLUSION
--Propsed --e-SBR
"0
�
'6
�
2C
0...
ct?i
I ---+--I � 3>-'1 -
::L4:\ I '*::::"
����;"'::::hb�� ::: ��:::=::::::/;.. :r:� �" �.
'0 0 z
:--...
2
6
4
8
No. of Scenario s
---.A.-
Radiosity
sented to develop a new indoor propagation
Near field
prediction algorithm including rough surface
KA-MoM
--e-- BT
--*-
In this paper,an integrated technique is pre
BDPT
10
scattering. The BSP is used in combination with the ray tracing for the acceleration pur pose. This decreases the prediction time and also increases the accuracy. For better accu racy,rough surface scattering is introduced, which is analyzed by Oren-Nayar theory and
Fig.S Comparison in terms o/number o/predicted rays
forward-backward function of objects and walls. A concept of ICS is introduced for
450"--
-.
�Q)
E i=
.-
--------
-.
------
-.,,-------� -- Proposed
------
�
400
3
-.
------
I
I
�..
-=-017, ..
II--e-SBR
1',
50
3
00
25
2
6 No. of Scenarios
4
increasing the accuracy of the proposed algo
8
rithm, while NOP technique is presented to reduce the intersection test time. The obtained
---+---
Radiosily
results from this study show that the proposed
�N,a'fi'"
algorithm achieves as a maximum 37.83%
---.A.-
KA-MoM
--e--
---*-
BT BOPT
10
Fig.6 Comparison in terms 0/computation time
higher accuracy and 34.44% lower time con sumption than that of the existing algorithms. The proposed hybrid technique can not only be used in radio signal prediction but also it can be used for target identification,remote sensing,or in particular areas in electromag netic research where rough surface scattering is crucial. Although scattering is included in the pro
ter, which has influence on the electric field strength. For different types of antenna, the results will be different. Figure 9 shows the electric field strength in the receiver side for half wave dipole antenna with 2.15 dBi gain and parabolic antenna with 18 dBi gain. In all cases, 2.4 GHz operating frequency is used
155
posed algorithm for propagation prediction, this can be also used in inverse scattering for determining object characteristics,such as its shape,internal construction,etc. based on data of how it scatters the incoming radiation or particles. In this paper,we have proposed ray tracing technique for indoor environments. We China Communications· October 2014
cannot directly use this technique for outdoor
Table II Results for all five environments
environments or corridors. Hence, there is a great chance of improvement of this algorithm in the future. Our next target is to gear up the
SBR
existing algorithm for inverse scattering and
Radiosity
then,we will make this algorithm a universal
Near field
ray tracing tool for using in any kind of envi
KA-MoM
ronments. ACKNOWLEDGEMENT
BT BDPT Proposed
2
Time (ms) 3
295
231
258
3 91
283
371
248
328
345
250
329
243
260 236
Our sincere thanks go to the University of
291
225
312
221
201
224
203
218
301
References
E i=
HG, Siwiak K, Win Ml. A Comprehensive Stan
80
60
dardized Model for Ultrawideband Propagation Channels. IEEE Transactions on Antennas and
2
v
Propagation 2006; 54(11):3151-3166. Pan XM, Cai L, Sheng XQ. An efficient high (/)
ro
order multilevel fast multipole algorithm for
213
172
233
139
106
154
146
131
147 117
94
152
158
168
156
181
194
[3 .0
_._
._. .
--
--'
4 6 No. of Scenarios
3
93
4
113
142
136
178
114
146
79
167
102
138
139
165
5
179 143
165
187
182
203
161
199
SF=8 SF=4
_._
._..
--
--'
t5 '6 !!!
ray tracing techniques using ray frustums for wireless propagation models. Progress In Elec
[3 .0
--
.Ql 350
Kim H, Lee HS. Accelerated three dimensional
8
10
(a)
400
(/) '0 Q)
Electromagnetics Research 2012; 126:85-100.
SF=20 SF=8 SF=4
a..
'0 250
tromagnetics Research 2009; 96:21-36.
0
Poli L, Rocca P. Exploitation of T E-TM scatter
z
ing data for microwave imaging through the multi-scaling reconstruction strategy. Progress
200
2
In Electromagnetics Research 2009; 99:245-260. [5]
254
184
215
122
c
electromagnetic scattering analysis. Progress In
[4]
193
204
2
Q)
Molisch AF, Cassioli D, Chia-Chin C, Emami S, Fort A, Kannan B, Karedal J, Kunisch J, Schantz
[3]
236
146
No. ofP"edicted Rays
t-..,
Grant (UMRG) scheme (RG098/12ICT).
[2]
193
5
160 rl--,-----r---�--r====, --SF=20
Malaya for offering extensive financial sup port under the University of Malaya Research
[1]
4
Hasar Uc. A new microwave method based on
4 6 No. of Scenarios
8
10
(b)
transmission scattering parameter measure ments for simultaneous broadband and stable permittivity and permeability determination.
Fig. 7 Effects of scattering factor
Progress In Electromagnetics Research 2009; 93:161-176. [6]
Pinel N, Johnson JT, Bourlier C. A Geometrical From a Rough Layer With Two Rough Surfaces.
Dielectric Rough Surfaces. IEEE Transactions on
2010; 58(3):809-816.
Antennas and Propagation 2011; 59(1):180-188. [10] Hongzhen C, Yueting l, Hongqi W, Chibiao D.
Tarng J H, Liu WS, Huang Y F, Huang JM. A
SAR Imaging Simulation for Urban Structures
novel and efficient hybrid model of radio mul
Based on Analytical Models. IEEE Geoscience
tipath-fading channels in indoor environments.
and Remote Sensing Letters 2012; 9(6): 1127-
IEEE Transactions on Antennas and Propagation
1131.
2003; 51(3):585-594. [8]
Bin L, lengyuan L, Yang D. A Fast Numerical Method for Electromagnetic Scattering From
IEEE Transactions on Antennas and Propagation [7]
[9]
Optics Model of Three Dimensional Scattering
[11] Mani F, Oestges C. A Ray Based Method to
Li lX. Numerical simulation bistatic scattering
Evaluate Scattering by Vegetation Elements.
from three-dimensional conducting rough sur
face with forward & backward method-based
IEEE Transactions on Antennas and Propagation
vector basic functions. lET Microwaves, Anten
[12] De Rousiers C, Bousseau A, Subr K, Holzschuch
nas & Propagation 2010; 4(1):54-61.
China Communications· October 2014
2012; 60(8): 4006-4009. N, Ramamoorthi R. Real-T ime Rendering of
156
250..-,
Rough Refraction. IEEE Transactions on Visual
---r---'---�- I-I
ization and Computer Graphics 2012; 18(10): 1591-1602. [13] Junwen D, Va-Qiu J. Scattering Simulation and
Vi'
Reconstruction of a 3-D Complex Target Using
E
'Q;' 1501 � E
i=
� �
r
,I
Downward-Looking Step-Frequency Radar. IEEE
" x/ I ,p...,"1,1
Transactions on Geoscience and Remote Sens ing 2011; 49(10): 4035-4047.
-- Scattering Only
100 50 I'
[14] Ouattara V B, Mostarshedi S, Richalot E, Wiart J,
-e- Scattering+BSP
Picon O. Near- and Far-Field Models for Scat
--e- Scattering+BSP +NOP , 4
2
6
tering Analysis of Buildings in Wireless Commu
8
No. of Scenarios
10
nications. IEEE Transactions on Antennas and Propagation 2011; 59(11): 4229-4238. [15] Tao V B, Lee H, Bao HJ. KD-tree based fast ray
(a)
tracing for RCS prediction. Progress In Electro
en
magnetics Research 2008; 81:329-34l.
ro
c OJ
[16] Cocheril V, Vauzelle R. A new ray-tracing based wave propagation model including rough sur
US
"0
faces scattering. Progress In Electromagnetics
2
Research 2007; 75:357-381.
()
'5
[17] T hiel M, Sarabandi K. A Hybrid Method for
0..
�
Indoor Wave Propagation Modeling. IEEE Trans
-- Scattering Only
'0 ci z
actions on Antennas and Propagation 2008;
-e- Scattering+NOP
56(8):2703-2709.
[18] Franek 0, Andersen JB, Pedersen GF. Diffuse
II ==e=,,;;; Sc;;;a,;;tte ;; r;;;ing,g+�N ;; O � P=+�1 C�S=' L...----'-_____J
100 II�
4
2
6
8
No. of Scenarios
10
Scattering Model of Indoor Wideband Propaga tion. IEEE Transactions on Antennas and Propa gation 2011; 59(8):3006-3012.
(b)
[19] Zhang XV, Sheng XQ. Highly efficient hybrid method for monostatic scattering by objects on
a rough surface. lET Microwaves, Antennas &
Fig.8 Effects of scattering, NOp, and ICS on (a) Time and (b) No. of received valid
signals
Propagation 2010; 4(10):1597-1604.
E .......
G ., 41
... ,II 'JI!!Lln ' �" J
CJ)
U)
�
"0
�
31
� ' j\j
() 2r>