Building and Environment 49 (2012) 67e85

Contents lists available at SciVerse ScienceDirect

Building and Environment

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

Computational models for measuring spatial quality of interior design in virtual

environment
Aswin Indraprasthaa, *, Michihiko Shinozakib,1
a
b

Department of Engineering and Design, Shibaura Institute of Technology, Shibaura 3-9-14, Minato-ku, Tokyo 108-8548, Japan
Shibaura 3-9-14, Minato-ku, Tokyo 108-8548, Japan

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 20 April 2011
Received in revised form
9 September 2011
Accepted 15 September 2011

This paper presents a computational model designed to analyze and to assess quality of architectural
space. The model consists of two parts: rst part is a model of subdivided enclosed spaces, which is an
approximation of spatial layout regarding the enclosure and the circulation path. Second part is a model
of spatial quality assessment using three spatial parameters and two distinct approaches. The rst
approach of this assessment is visual distance and the second approach is viewing angle. The assessment
valued by these approaches then combined to obtain spatial quality ranking of each of the subdivided
enclosed space. Previous studies on spatial assessment showed the relationship between visual distance
and spatial quality can be modeled through mathematical approaches. Our work proposes an
improvement on the method of spatial mapping model and spatial quality assessment. Experiments have
been conducted on interior design and we developed spatial evaluation using three parameters: visual
openness, privacy and physical accessibility. Furthermore, we conducted a comparison study of privacy
assessment on design variances. Finding shows some distinctive results on the assessment approaches
that can lead to more elaborative spatial quality evaluation. The outcome on spatial quality assessment
can facilitate spatial quality evaluation of interior design in early stages of design development.

2011 Elsevier Ltd. All rights reserved.

Keywords:
Computational model
Interior space
Spatial qualities rankings
Virtual environment

1. Introduction
This study focused on the architectural space of interior design
as a result of the arrangement of architectural elements. An architectural space is dened as the void between physical boundaries of
the enclosures where its existence is independent of the users
presence. As an architectural space composed by its physical
setting, we developed a model using superimpose technique to
map this space into what we named as subdivided enclosed space.
The mapping procedure follows the basic relationships between
design elements and circulation paths using territorial lines
approach as studied by Koile [1]. The benet of this model is it
offers a more elaborative object of assessments specically relates
with spatial quality parameters. In previous study of spatial quality
evaluation [2,3], a single proxy usually used to represent object of
evaluation (i.e. center of the room or any arbitrary point in the
room). Some of earlier models have focus on spatial mapping

mechanism and procedures [4] with less emphasize on the development of spatial quality evaluation.
Our spatial mapping model results in an array of points in an
interior plan where each point has different spatial quality ranking
that related with their relative position to the architectural
elements. This model proposes a better analytic tool for spatial
quality evaluation. The comparison of our model with previous
works is presented in Table 1.
The spatial quality parameters used in this study are determined
and intended to improve previous achievements. For example,
Fischer-Gewirtzman and Wagner [5] and Pinsly et al. [6] analyzed
spatial openness and visual exposure. Both parameters related with
visual openness, which is bound for metric-based evaluation.
Demirkan et al. [7] used distance measurement to analyze privacy
in an interior space. We determine three spatial parameters for this
study: visual openness, privacy and physical accessibility.
2. Architectural spatial quality

* Corresponding author. Tel.: 81 3 6722 2732; fax: 81 3 5859 9249.

E-mail addresses: aswin@aswinindra.com (A. Indraprastha), sinozaki@sic.
shibaura-it.ac.jp (M. Shinozaki).
1
Tel.: 81 3 6722 2732; fax: 81 3 5859 9249.
0360-1323/$ e see front matter
2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.buildenv.2011.09.017

Architecture is experienced not just by attributes of its boundaries. The variability in the interior (enclosed space) and exterior
(enclosure) comprise the essence of architecture. Several architects
[8e10] as well as psychologists [11] have found that the

68

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

Table 1
Comparison of models and methods.
Gross method (1977)

Koile method (2001)

Sora method (2008)

Our method (2010)

Goal

Classication of enclosed space

by its boundaries

Quality of space by viewpoint

and geometric elements

Limitation

Only physical boundary forms

a subdivided enclosed space,
failed to accommodate center space, and no
hierarchical attributes related to the space
Enclosed space classication
based on its boundary

Design representation
by hierarchical
territorial space
Limit to the territorial
spaces as a result
of element, edge, and
circulation model
Abstraction model of
design representation

Only to measure quality

at arbitrary viewpoint
position relative
to the boundary element
Spatial quality measurement
based on viewpoint

Spatial map that represents

relationship of boundary
elements and circulation space
Experimented on
rectangular-based plan

Numerical values that represents

rank of enclosed space

Graphic representation
of model

Numerical result of
measurement

State of the art

Result/Output

experiences in the same space can vary with the changes in

architectural design elements such as colors and transparency.
Another aspect of architectural space is that architects always
compose space according to the boundaries and inter-relationships
between the architectural elements and its planned activities. March
and Steadman [3] used network graph to denote topological relationships between rooms e which room gives access to another,
which room is adjacent to another. Pna [4] suggested that the
matrix relationship of activities should be constructed to model
adjacency of spaces regarding designated activities. Both studies had
used basic mathematical model of spatial conguration in order to
understand relationship between spatial design and activities.
Our model of enclosed spaces is the elaboration from previous
study with the advancement of having a map conguration of what
we named as subdivided enclosed space inside a set of interior
space. We advice each area of subdivided enclosed space expresses
distinct spatial quality that is signicant for spatial quality analysis
and further conceptual design elaboration.
3. Mapping the interior by virtual agent
The concept of spatial quality mapping by previous works
referred to the preposition that on an enclosed space or interior
space, the hierarchy of the space can be mapped by the strength of
inuence of its boundary elements [12e15]. A study of spatial
mapping has been conducted by using number of solid boundaries
to classify enclosed spaces [13], combining concept of territory
space to map strength inuence of boundary elements into
enclosed space [1] and recently, used distance model to classify
hierarchy of enclosed space [14].
Our model began with the classication of enclosed space [15]
where there are six dened architectural spaces: by linear
elements (column), by single vertical plane (i.e. wall), L-shaped
space, parallel space, U-shaped space and four planes to dene full
closure of space. The basic enclosed space category we used for this
research is L-shaped space (Fig. 1).

Spatial map based on hierarchical

relationship of boundary
elements and circulation space
Layout of spatial mapping
resulting in determination of
subdivided enclosed space

Following the basic classication of enclosed space, our model of

spatial mapping is based on the visual perception and position of
the virtual agent. Therefore, the rules for generating spatial
mapping are taking account the detected objects by virtual agent.
Our method of enclosed space detection is based on the hierarchical structure of the space as previously been studied. On this
model, the hierarchy of enclosed space is as follows:
1. L-shaped space; dened by two non-collinear solid boundaries.
2. Circulation space; dened where it has territorial space of
circulation gate (i.e. door).
3. Attractor space; dened where it has territorial space of other
architectural elements at the boundary such as openings.
4. Subdivided space; dened as result of the rules of spatial
mapping.
As result of this concept, we determined an interior space as
a unit analysis. An interior space is a space that comprises of
enclosures, enclosed space and circulation path. We named this
area as bounded area (Fig. 2).
A bounded area comprises of rectangular-type arrangement of
enclosed spaces that depend on the conguration of architectural
elements at its enclosure. We employed procedural rules by
considering relationship between bounded area and its circulation
path. The bounded area may enclosed by opaque boundaries (walls),
openings (windows or doors) and other architectural elements. A
circulation path is a circulation connector inside the bounded areas.
Circulation gates such as door and path form a circulation space
inside the bounded area.

Architectural plan

Interior space(s)

Enclosed space

Enclosure
Circulation path

Bounded area
for computational analysis
Fig. 1. Diagrammatic model of enclosed space, a) L-shaped space; b) U-shaped space.

Fig. 2. Unit of analysis.

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

69

Variant of gate position

Ed
Md

Enclosed space by center point

Enclosed space subdivided by rules

1

2
3

3
2

2
1

Fig. 3. Subdivided enclosed space having a path terminated in the space.

a Variant of gates position

b Enclosed space by center point

c Enclosed space subdivided by rules

1

2
3
2
1

2
2

3
2
1

2
1

3
2

1
2

3
2

2
1

Fig. 4. Subdivided enclosed space having a linear path passing through space.

2
3

70

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

For determining subdivided enclosed space, our computational

method works by analyzing set of points detected by agent as he
loosk around his surroundings. This analysis comprises of procedural rules based on spatial hierarchy of bounded area.

Variant of gates position

3.1. Model of circulation path in a bounded area

Enclosed space by center point

Enclosed space subdivided by rules

As previously mentioned, procedural rules for creating a map of

spatial conguration is based on the relationship between bounded
area and its circulation path. The computation for dividing enclosed
space is conducted under the rules of circulation path inside the
bounded area. Here, the pathespace relationship is dened as the
connection between bounded areas and circulation paths.
In architecture, a circulation path is the connection in which
people traverse from one node to another [1,4]. In narrower scope,
a circulation path is the connection between a circulation gate
(door) and the space within the enclosure. There are two ways
considered as the relationship between the paths and space [15]:
terminate in space and pass through space.
3.1.1. Terminate in space
If a room has only one door, then the circulation path is terminated inside the room. The model of subdivided enclosed space by
circulation path is then the result of the relationship between the
enclosed space and circulation area.
Fig. 3 illustrates procedural steps of dividing enclosed space
having a circulation path terminated in the bounded area.
The rules to subdivide enclosed space are as follows:

Final result
3

2 2
3

2 1

2 1
1

2
1

2
2

3
2
1

3
1

3
2

2
1

Fig. 5. Subdivided enclosed space having angular path passing through space.

 Determine L-shaped enclosed space (number 1 in Fig. 3c) by

the center point of each boundary (Fig. 3b);
 Determine remaining space (number 2 in Fig. 3c) with respect
to the circulation gate and center point;
 Determine the axial line along the center point and perpendicular to the path direction;
 Make parallel lines from the gate to the axial line (Fig. 3c) to
obtain circulation space (number 3 in Fig. 3c).
Following these rules, we established conguration of enclosed
space having a circulation space that terminates in the center of the
interior space.

Fig. 6. Set of points based on x-axial line and y-axial line.

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

71

Fig. 7. Illustration on determination of subdivided enclosed space according to x-axial line and y-axial line.

3.1.2. Pass through space

If there is more than one circulation gate in a room, the relationship between the paths and space known as pass through
space. We developed two models: cut axially and cut obliquely. In
both models, the circulation paths may have linear or non-linear
positions.

3.1.2.1. Linear path. The circulation path is dened as linear if two

gates are connected by a straight line. The rules to determine
subdivided enclosed space are the same as the previous rules. Fig. 4
illustrates procedural steps of dividing an enclosed space having
a linear circulation path passing through a bounded area.
The rules of this relationship are extensions of the rst set of rules.

72

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

c
b
a

Area of windows=0.64 (m2)

Area of wall=26.125 (m2)
Ratio=0.02

Area of windows=1.16 (m2)

Area of wall=26.125 (m2)
Ratio=0.04

Area of windows=3.72 (m2)

Area of wall=26.125 (m2)
Ratio=0.142

Fig. 8. Determination of center point of each subdivided enclosed space based on

quadrant position.

3.1.2.2. Angular path. The path is dened as angular if two gates are
connected by angular line. In this situation, we hypothesized that
the center point is a signicant factor to determine subdivided
enclosed spaces and circulation spaces. By considering this, when
two paths encounter a space, their connection must pass through
a center point or an axial line as dened by the previous rules.

Fig. 10. The size of window relative to their attached wall determines value of
transparency.

Fig. 5 illustrates procedural steps of subdividing an enclosed

space having angular circulation path passing through a bounded
area.

Fig. 9. Graphical example of computation of subdivided enclosed space, a) Initial plan; b) determination of L-shaped enclosed space; c) circulation area using axial line; d)
determination area of opening territory by extending its edges perpendicular to axial line; e) determination of extended area by extending L-shaped edges perpendicular to the
nearest lines; f) center points of the subdivided enclosed spaces.

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

73

Table 2
Two approaches to measure visual openness.
By distance

By viewing angle

Average distance
Average distance from q to
from p to window 1 and window 2 X1
window 1 and window 2 X2
X1 < X2
Visual openness at p is lower than visual openness at q

Ratio of viewing angle and

Ratio of viewing angle and
maximum angle at p Y1
maximum angle at q Y2
Y1 < Y2
Visual openness at p is lower than visual openness at q

The rules to subdivide enclosed space are as follows:

 Dene L-shaped enclosed space by the center point of each
boundary (number 1 in Fig. 5c);
 Evaluate remaining space with respect to the circulation gate
and center point (number 2 in Fig. 5c);
 Determine the minimum convex area;
 Divide the area by its center point;
 Determine axial lines along the center point and perpendicular
to the path direction;
 Determine parallel lines from each gate to the axial line
(number 3 in Fig. 5c).
3.2. Algorithm and procedural rules
The followings are rules that are executed sequentially over set
of points. These rules are applied following results of object
detection and inner point computation over the inner coordinates
of boundary objects relative to the position of agent.
The general procedures are as follows:
1. Determine the center point of the polygon c (xc,yc) by nding
out the minimum and maximum point of the vertices.
2. Determine the horizontal axial line and the vertical axial line.
From the previous step, y yc horizontal axial line;
x xc vertical axial line.
3. Dene a set of points based on the two axial lines. This is done
by sorting the points along the horizontal axial line, followed
by sorting the points along the vertical axial line.
3.1. Sorting the points along the horizontal axial line is done by
dividing the points into two sets: those that are located
above the line (ypi> yc), and those that are located below

the line (ypi < yc). They are dened as xpi set of points
above the x-axial line and xpi set of points below the xaxial line:
x
pi (xxpi,ypi) where ypi > yc; xpi (xxpi,ypi) where
ypi < yc.
3.2. Accordingly, sorting the points along the vertical axial line
is done by dividing the points into two sets: those that are
located on the right side of the line (xpi > xc), and those that
are located on the left side of the line (xpi < xc). They are
dened as ypi set of points on the right side of the yaxial line and ypi set of points on the left side of the yaxial line (Fig. 6)
y
pi (xpi,yypi) where xpi > xc; ypi (yxpi,ypi) where
xpi < xc.
4. Determine the intersected points at the horizontal and vertical
axes. As of now, we have four sets of points: {xpi}, {xpi}, {ypi},
{ypi}. Each of these set members must be mapped into the
designated axis, so that we have intersected point sets.
5. Determine the midpoint of two sequential points from each
intersected point set.
6. Determine the center point of the result of Step 5.
Fig. 7 illustrates the overall procedure as explained above. The
procedure for combining intersected points on both the axial lines
results in generating divided enclosed spaces. The next stage is to
generate the center points of these spaces by running procedural
rules on each point. We determined these center points based on the
evaluation of the intersected points of each line by their quadrant.
As previously explained, we separated each detected point
according to their quadrant with respect to the axial lines, therefore
we obtained four sets of points: {x}, {x}, {y}, and {y}. The
center point of the subdivided enclosed space is given by means of

Fig. 11. Cases to determine maximum viewing angle at p.

74

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

window
1

4
6

door

2
5

10

14

15

11

door

12

13

16

17
door

18 19
20

21

22

26

27

23

24 25
window
0.5

1. 0

Fig. 12. Result of visual openness measurement on room 1 of Kaufmann house.

their position in the quadrant, which is determined as the relationship of these sets: {x, y}, {x, y}, {x, y}, {x, y}.
Fig. 8 below illustrates this concept:
This approach maintains the origin of the intersected points at
the axial lines and results in a better calculation for determining
center point of each subdivided enclosed space.

3.2.1. Determine axial lines

The determination of axial lines is signicant in our model as
a guideline for subsequent procedures. The axial lines are determined by determining the area of a polygon based on the set of its
boundary points and computing its center points at the x-axis and
at the y-axis. The necessary equations are:

Fig. 13. Kaufmann House by Frank Lloyd Wright.

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

75

Fig. 14. Rachofsky house by Richard Meier.

i1
1X
Area of boundaryA
x y
 xi1 yi
2 i 0 i i1
i1
1 X
Center at X
x y
 xi1 yi xi xi1
6A i 0 i i1

(1)

(2)

Center at Y

i1
1 X
x y
 xi1 yi yi yi1
6A i 0 i i1

(3)

3.2.2. Intersected points at axial lines

Dene the axial lines as: y yc and x xc, and (xc,yc) is the center
point. At this stage, we evaluate the points of each quadrant with
the axial lines (refer to Fig. 5):

76

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

Table 3
Procedures of measurement for spatial quality level.

Procedures for measurements

VO: visual openness

PR: privacy

Calculate average distance

Pn
vo
i 1 dWi
Dp
n

Dip

Pk
pr

i1

Dp
Dp

PRp Exp

i1

pr

dWip

pr 2
p
pr
Dp

Calculate level using distance method by factor of transparency index

VOD VOD tr
PRD PRD tr
p
p

Pn

ac

Dp

i1

dDac
ip

ACp Exp

ac 2
p

D

ac
Dp

ACD ACD tr
p

Calculate ratio of covered angle

uvo
p
5

Pk

pr

dWip

Calculate strength of inuence

VOD Exp

AC: physical accessibility

qp
100

uac
p

qp
100

upr
p

qp
100

Calculate strength of inuence

VOup Exp

u2p
up

PRup Exp

upr 2p

upr
p

Normalize value of (3) and (5)

Combine both result and obtain arithmetic average

1
1
VOp VOD VOup
PRp PRD PRup
p
p
2
2

 Quadrant 1 {(xpi,ypi)} where xpi < xc and ypi > yc; intersected
points at y; yc {(xpi,yc)}; intersected points at x; xc {(xc,ypi)}.
Set points in quadrant 1 {q1(xpi,yc) and q1(xc,ypi)}.
 Quadrant 2 {(xpi,ypi)} where xpi > xc and ypi > yc; intersected
points at y; yc {(xpi,yc)}; intersected points at x; xc {(xc,ypi)}.
Set points in quadrant 2 {q2(xpi,yc) and q2(xc,ypi)}.
 Quadrant 3 {(xpi,ypi)} where xpi < xc and ypi < yc; intersected
points at y; yc {(xpi,yc)}; intersected points at x; xc {(xc,ypi)}.
Set points in quadrant 3 {q3(xpi,yc) and q3(xc,ypi)}.
 Quadrant 4 {(xpi,ypi)} where xpi > xc and ypi < yc; intersected
points at y; yc {(xpi,yc)}; intersected points at x; xc {(xc,ypi)}.
Set points in quadrant 4 {q4(xpi,yc) and q4(xc,ypi)}.
3.2.3. Midpoints at each set in the quadrant
For each set, the program computes the midpoints on two
sequential points of the sorted set. The denition and rule is as follows.
Denition and rule:
 {qn(xpi,yc)} and {qn(xc,ypi)}; n (1,.,4); set of points at quadrant
n by horizontal and vertical axes.
0
0
 {qn (xpi,yc) and qn (xc,ypi)}; n (1,.,4); set of sorted points at
quadrant n by horizontal and vertical axes.

ACp Exp

ACp

ac 2
p

D

ac
Dp

1
ACD ACup
p
2

 {Mid_qn(x)(xpi,yc)} midpoint set at quadrant n by horizontal

axis.
 {Mid-qn(y)(xc,ypi)} midpoint set at quadrant n by vertical axis.
0
{Mid-qn(x)(xpi,yc)} {qn (((xpi xpi1)/2),yc)}
Mid-qn(y)
qn0
(xc,ypi)} { (xc,((ypi ypi1)/2))}
{

3.2.4. Center points at each set in the quadrant

The center points of each quadrant are determined by
combining the midpoints at the vertical and horizontal lines.
The denition and rule is as follows:
Denition and rule:
 {Mid_qn(x)(xpi,yc)} midpoint set at quadrant n by horizontal axis.
 {Mid-qn(y)(xc,ypi)} midpoint set at quadrant n by vertical axis.
 {Cen_qn(pi)} center points set at quadrant n.
{Cen_qn(pi)} {Mid_qn(x)xpi,Mid-qn(y)ypi}.
All computational process conducted in Virtools. The result of
this process is the conguration of subdivided enclosed spaces as
illustrated in Fig. 9:

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

77

Table 4
Visual openness index measurement on house plans.
Kaufmann house plan

Two windows

Rachofsky house plan

One window

One window

4. Model of spatial quality parameters

In an architectural spatial quality model, we proposed factors
that inuenced enclosed space quality which are visual openness,
privacy and physical accessibility. These basic parameters could
have effect on how we regard particular interior space or enclosed
space based on the conguration of architectural elements of
enclosure. The previous work showed an empirical nding that the
viewing distance and viewing angle yielded signicant inuence
toward spatial evaluation and orientation [16]. Therefore, we
develop our method of measurement based on the distance of
architectural element to the point of measurement and the
covering viewing angle from point of measurement to particular
architectural element.

One window

One window

The result of this model is numerical level attached to each of

the subdivided enclosed space that gives ranking on the particular
spatial parameter.
The followings are explanation of each parameter:
4.1. Visual openness parameter
In modeling visual openness, we considered window as the
source of its inuence. This visual openness inuence is received by
each of subdivided enclosed space in a different level regarding
their layout in a bounded area. The signicance of this index is that
it can express the visual openness inuence upon enclosed space
with such qualitative and quantitative attribute attached such as
privacy, natural light and air.

Table 5
Privacy index measurement on house plans.
Kaufmann house plan

Two windows, three doors

Rachofsky house plan

One window, two doors

One window, one door

One window, two doors

Two doors

One window, three doors

78

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

Table 6
Physical accessibility index measurement on house plans.
Kaufmann house plan

Three doors

Rachofsky house plan

Two doors

One door

The term visual openness index refers to the level of visual

inuence at a center point of a subdivided enclosed space. We
dened three variables to compute visual openness:
1. Visual distance; refers to the distance from a reference point p
to the openings (i.e. windows);
2. Transparency ratio; refers to the ratio of opening area and its
adjacent wall area;
3. Viewing area; refers to the ratio of viewing area from point p
having all windows and maximum viewing area dened as 100 .
We dened transparency level here is a value drawn by fraction
of the dimension between opening and remaining solid surface
where the opening is located.
Fig. 10 illustrates this concept:
The procedure to compute visual openness is by taking average
value of visual openness by visual distance and transparency ratio
and visual openness by viewing area. Table 2 shows our approach to
combine measurements of visual openness:
4.2. Privacy parameter
Privacy parameter or visual privacy parameter refers to the
visual penetration of a point p in bounded area as a result of being
viewed from external spaces [6,7]. In our model, level of privacy
depends on two factors: openings (i.e. window) and circulation
gate (i.e. door).
Aside of the function to circulate light and fresh air; window is
an architectural element that provides visual exchange by its level
of transparency. We used same approach to determine transparency ratio as with the visual openness parameter.
The procedure to compute privacy is by taking average value of
privacy by distance and privacy by viewing area. By distance
method, the greater distance from window and/or door, the greater
privacy level. Accordingly, greater angle to cover all windows and/
or doors is resulting in lower privacy level.
4.3. Physical accessibility parameter
In our model, physical accessibility parameter refers to the value
of average distance from point p at the center of a particular subdivided enclosed space, to the doors in that bounded area. The

Two doors

Two doors

Three doors

physical accessibility index represents ease of accessibility of

a subdivided enclosed space in an interior space.
The procedure to compute accessibility is similar with the
previous where using two approaches, except for computing accessibility index, the object of measurement is circulation gate (i.e. door).
4.4. Procedures of measurement
Each of spatial quality parameter has distinctive characteristic
regarding the inuential strength of its source, for example:
1. Visual openness parameter at p in space; the greater average
distance to the windows means the lower visual openness
index and the greater fraction of viewing angle means the
greater visual openness index.
2. Privacy parameter at p in space; the greater average distance to
the windows and the doors means the higher privacy index and
the greater fraction of viewing angle means the lower privacy
index.
3. Accessibility index at p in space; the greater average distance
to the doors means the lower accessibility index and the
greater fraction of viewing angle means the greater accessibility index.
The illustrative diagram of method of measurement for visual
openness, privacy and accessibility index is presented in Table 3.
4.5. Denition and equation
The procedure for computation takes some abbreviations and
denitions as follows:
dWp distance from p to center point of window;
dDp distance from p to center point of door;
aWi area of window i; aWLi area of adjacent wall of window i;
aDi area of door i; aWDi area of adjacent wall of door i;
n number of window; k number of doors;
trwi (aWi/aWLi) transparency index of window i; trdi (aDi/
aWDi)transparency index of door i;
qvo visual angle at p having all window covered; uvo visual
openness ratio of covered angle;

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

79

Fig. 15. Spatial qualities level on Kaufmann house.

qpr visual angle at p having all windows and doors covered;

upr privacy ratio of covered angle;
qac visual angle at p having all doors covered; uac physical
accessibility ratio of covered angle.

4.6. The viewing angle method

In addition to the distance method, we propose the viewing
angle method for measuring inuence of spatial parameters relative to the position of the point of reference.

80

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

Fig. 16. Spatial qualities level on Rachofsky house.

The viewing angle of measurement is obtained as a fraction of

angle u to cover outer edge of object of measurement to the
maximum model of viewing angle. We determined the model of
a maximum horizontal viewing angle as 100 .
The variable u centered at the point p and its direction vector
depends on the position and layout of the object of measurement.
The principle of this approach is to obtain maximum angle to cover

outer edges of object of measurement from p. This concept is

illustrated in Fig. 11:
4.7. Normalization
There are two numerical values as the results from two distinct
methods. To obtain average value at each point p, we made
normalization on each result as follows:

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

81

Fig. 17. Summary of spatial qualities evaluation on case studies.

Denition:
1. SP as spatial parameter, SP {VO, PR, AC};
2. SP dp as spatial quality index by distance at p; SP sp as
spatial quality index by viewing angle at p;
3. nSP dp as normalization by distance at p; nSP sp as
normalization by viewing angle at p;
4. MaxSP d as maximum value of spatial quality index by distance
of all points p in space; MinSP d as minimum value of spatial
quality index by distance of all points p in space;
5. MaxSP s as maximum value of spatial quality index by
viewing angle of all points p in space; MinSP s as minimum
value of spatial quality index by viewing angle of all points p
in space.


nSP dp

SPdp  MinSP d

MaxSP d  MinSP d

nSP d0; 1

(4)

at a particular point of reference. By this case, it is appropriate to

combine both results to give an average weight of spatial qualities
on each point of reference. We use arithmetic mean of these both
values to represent average value obtained by objective and
subjective variable.
5. Experiments and result
The input for the experiments is CAD polygon. We used the
architectural plans of Kaufmann House [17] (Falling Water house)
by Frank Lloyd Wright (Fig. 13) and Rachofsky house by Richard
Meier [18] (Fig. 14).
The experiments on six different layouts are explained as
follows (Tables 4e6). Each table below illustrates detailed layout of
the rooms, position of center point of subdivided enclosed spaces
and architectural elements characteristic (i.e. number and dimension of wall, window and door).

and


nSP sp


SP sp  MinSP s
;
MaxSP s  MinSP s

5.1. Result

nSP s0; 1

(5)

4.8. Combined value

On one example of visual openness measurement (Fig. 12), it is
showed that there are different results of visual openness level that
relates to the position of the point of reference to the window. By
their denitions, the distance and visual angle method has characteristic that may lead to how we quantify visual openness
objectively. By distance method, the visual openness level depends
on the average distance to all windows.
However, as suggested by Henry [16], visual distance also
inuenced by viewing area that may lead to subjective evaluation

Fig. 18. Comparison result.

The results of spatial quality measurements on six rooms are

presented as follows; rst is the spatial qualities measurement on
Kaufmann house and Rachofsky house (Figs 15 and 16). Second is
comparison on the result of both case studies. We presented the
spatial quality level of each reference point as the combined result
by distance method and by viewing angle method.
Based on this evaluation, spatial quality evaluation depends on
the layout conguration of the openings that made up the spatial
mapping as we developed. In such, the openings affect the result of
spatial quality ranking. The rooms of Kaufmann house serve as
bedrooms while room 1 and room 3 of Rachofsky house serve as
multipurpose rooms. Fig. 17 (a and b) presents the results of spatial
qualities of three parameters: visual openness, privacy and physical
accessibility considering the area of each room.
Fig. 17(a) presents a result that can be used for quantitative
spatial quality over the same function. In this case study, room 3 of
Kaufmann house has the highest level of visual openness, privacy
and physical accessibility. This result shows the relationships of the
subdivided enclosed space as reference of measurement, the layout
of the openings and spatial qualities. Although it needs more
empirical nding, it is interesting to note that this method gives
a possibility of more elaborate and objective analysis over spatial
qualities parameters.
The diagram (Fig. 18) shows that the physical accessibility and
privacy levels of three plans of Rachofsky house are lower than the
Kaufmann house. The shapes of the plan and conguration of the
openings contribute to this result. On the case of visual openness
evaluation, the dimension of the windows in Rachofsky house
contributes for increasing its level compared with the Kaufmann
house.

82

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

Table 7
Spatial qualities measurement on design alternatives.
Design variance

Spatial qualities level

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

83

Table 7 (continued )
Design variance

Spatial qualities level

Fig. 19. Spatial qualities optimization.

6. Application for design alternatives analysis

In this section, we demonstrated practicability of spatial qualities evaluation using previously explained methods. As a tool to
improve design process, this demonstration showed analytical
examination over series of design alternatives. On the case of visual

window
1

door

3
4

10

11

14

15

openness (VO), privacy (PR) and physical accessibility (AC)

parameters respectively, this analysis can help to improve spatial
quality of a bedroom design.
Regardless the unlimited possibility to create design alternatives, these cases demonstrated rational iteration of design
improvements by transformation (rotate, translate) of design
elements from original design. Here, each of design alternatives has
its nal layout by transformation of windows and doors with their
dimensions remain intact. On each alternative, the layout is
developed intuitively and thus, independent from the intention or
specic purpose for spatial qualities parameters.
Since the results of these efforts are to be compared with the
original design, all measurements results are valued relative to
original design.
The spatial qualities levels over design alternatives are presented as follows (Table 7):
By comparing each design alternatives with original design on
spatial quality parameters, improvements on spatial qualities can
be shown as following diagrams (Fig. 19):
All design variances have higher privacy level than the original.
However, alternative 5 has highest level of spatial qualities. In this
regards, with the variables of measurement are the position of
openings and circulation gate, we can optimize design plans based
on three spatial qualities parameters.

door

door

12

13

16

17
door

18 19
20

21

22

26

27

23

24 25

window

window
0 .5

0 .5

1. 0

Room 1

1. 0

R oom 3
Fig. 20. Visual openness level of the rooms in Kaufmann house.

10

84

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

Table 8
Visualization of original design and best design alternative.

Original design

3D view 1

3D view 2

Best improvement

3D view 1

3D view 2

7. Summary and discussion

7.2. On spatial qualities evaluation

7.1. On mapping interior space

The model of spatial qualities evaluation using three spatial

parameters has been experimented. Furthermore, this evaluation
using distance and viewing angle methods revealed the results that
had not been reviewed in previous studies. In our model, by
combining distance and viewing angle methods, we can manage to
compute inuential strength of an architectural element as it is
received at a particular point in space. The decay function has been
employed to give distinctive result over relative small variants
given the context of interior plan.
This result as well revealed the relationship between layouts of
the interior design and measurement variables that had some
interesting result (Table 8).
In this context, in addition to give visual cues, numerical values
of spatial qualities levels represent inuential strength of the
parameter on the area of subdivided enclosed space.
This model of spatial mapping gives opportunity for deeper
spatial analysis over the conguration of architectural space. Based
on the boundary elements and the relationship between space and
circulation path, this model can be developed further to analyse
design preferences of an architect by his spatial designs.

We presented a methodology for mapping architectural space

by developing a semi-automatic method to generate subdivided
enclosed spaces over CAD-type data. Our nding revealed the
improvement of spatial qualities analysis using subdivided
enclosed spaces as area of references. Moreover, generating map of
interior space facilitates a visual cue to help designer develop better
spatial conguration of his plan. This is based on the proposition
that our method uses the relationship between architectural
elements and circulation space that made up the subdivided
enclosed spaces. In example, the analysis of visual openness on
a series of design plan with the same function will facilitate
understanding on organizing architectural space in the process of
design as depicted in Fig. 20:
In Fig. 20 colors gradient indicates level of visual openness and
circle radius represents model of inuential area of the visual
openness level (For interpretation of the references to color in this
paragraph, the reader is referred to the web version of this article.).
Another possibility for the implementation of this method is to
analyze the spatial signature of particular architects or styles in
addition to the typological study based on architectural elements of
the enclosure.
On its limitation, this work remains focuses on rectangularbased interior plan with only use three physical setting variables:
wall, window and door. We are aware that for future improvement
there are some signicant factors need to be addressed such as
considering more architectural elements including height of the
ceiling.

References
[1] Koile Kimberle. The architects collaborator: toward intelligence design tools
for conceptual design. PhD Dissertation. MIT; 2001.
[2] Alexander Christopher, Ishikawa Sara,
Silverstein Murray. A pattern
language: Towns, Buildings, Construction. New York: Oxford University Press;
1977.
[3] March Lionel, Steadman Philip. The geometry of environment. Cambridge: The
MIT Press; 1971.

A. Indraprastha, M. Shinozaki / Building and Environment 49 (2012) 67e85

[4] Pena William. Problem seeking: an architectural programming primer. 4th ed.
Wiley; 2001.
[5] Fischer-Gewirtzman Dafna, Wagner Israel A. Spatial openness as a practical
metric for evaluating built-up environments. Environ Plann B Plann Des 2003;
30:37e49.
[6] Pinsly Dalit Shach, Fisher-Gewirtzman Dafna, Burt Michael. Visual Exposure
analysis model: a comparative evaluation of three case studies. Urban Des Int
2007;12:155e68. Palgrave Macmillan.
[7] Demirkan H, Osman Demirbas O. Privacy dimension: a case study in the
interior architecture design studio. J Environ Psychol 2000;20:53e64.
[8] McGlynn Sue, Smith Graham, Alcock Alan, Murrain Paul. Responsive environment. In: Ian Bentley, editor. Oxford: Architectural Press; 1985.
[9] Rasmussen Steen Eiler. Experiencing architecture. Cambridge: MIT Press; 1959.
[10] Lawson Bryan. The language of space. 5th ed. Architectural Press; 2007.
[11] Tversky Barbara. Three dimensions of spatial cognition. In: Theories of
memory, vol. II. Psychology Press; 1998.

85

[12] Leupen Bernard, Grafe Christoph, Korning Nicola, Lampe Marc,

Dezeeuw Peter. Design and analysis. New York: Van Nostrand Reinhold; 1997.
[13] Do Ellen Yi-Luen, Gross MD. Tools for visual and spatial analysis of CAD
models. In: Computer assisted architectural design futures 97. Kluwer
Academic Publisher; 1997. pp. 189e202.
[14] Key Sora, Gross Mark D, Do Ellen Yi-Luen. Computing spatial qualities for
architecture. In: Proceeding of ACADIA 2008; 2008. p. 472e7.
[15] Ching Francis DK. Architecture: form, space and order. 3rd ed.. New Jersey:
John Wiley and Sons; 2007.
[16] Henry Daniel. Spatial perception in virtual environment: evaluating an
architectural application. In: Virtual reality annual international symposium,
IEEE Xplore; 1993. p. 33e40.
[17] Hoesly Bernhard. Frank Lloyd Wright: Fallingwater, AU 118; July 1980.
p.155e66.
[18] Meier Richard, Futagawa Yukio, Futagawa Yoshio. Richard Meier, GA document, vol. 8; 2008.