Human Task and Disability Based Automated Evaluation of Space and Design in CAD

Abstract

As data continues to influence every field, including creative ones such as architecture and design, the importance in the type of data increases. Through generative algorithms, or more recently machine learning, the mathematical formulas are designed to reduce, minimize, or optimize certain parameters. These parameters should have a method for evaluation in order to check the generated or created design option. It is this method of evaluating the design that is a critical underlying component that must be researched before implementing mass customized designs from computer algorithms. This paper discuses the use of human factors as an evaluation criteria, with the important assumption: the designed space is meant to be for people. A broad review of various human factors that have been quantified in literature is presented. Additionally , methods for parsing a building design when geometry is not organized, as in the case of BIM, are detailed. Examples of human factors related to designed spaces of various people are given, along with specifics in how the presented evaluation methods can be applied to improving space for people, including some with disabilities.

Full text

181
Human Task and Disability Based Automated Evaluation of
Space and Design in CAD
Mathew Schwartz1, Adara Azeez1and Kanisha Patel1
1CoAD, New Jersey Institute of Technology, Newark, USA, cadop/ala33/kp273@njit.edu
ABSTRACT
As data continues to influence every field, including creative
ones such as architecture and design, the importance in the
type of data increases. Through generative algorithms, or
more recently machine learning, the mathematical formulas
are designed to reduce, minimize, or optimize certain param-
eters. These parameters should have a method for evaluation
in order to check the generated or created design option. It is
this method of evaluating the design that is a critical underly-
ing component that must be researched before implementing
mass customized designs from computer algorithms. This pa-
per discuses the use of human factors as an evaluation criteria,
with the important assumption: the designed space is meant
to be for people. A broad review of various human factors
that have been quantified in literature is presented. Addition-
ally, methods for parsing a building design when geometry is
not organized, as in the case of BIM, are detailed. Examples
of human factors related to designed spaces of various people
are given, along with specifics in how the presented evalua-
tion methods can be applied to improving space for people,
including some with disabilities.
Author Keywords
Python; Human Factors; Simulation; Evaluation; Graph;
Disability;
ACM Classification Keywords
I.6 SIMULATION AND MODELING : [General]; J.5 ARTS
AND HUMANITIES: [Architecture]; D.2.2 SOFTWARE
ENGINEERING: Design Tools and Techniques—Human
Factors; I.3.6 COMPUTER GRAPHICS: Methodology and
Techniques—Ergonomics
1 INTRODUCTION
A fundamental question must be asked when designing a
space: is the space for people? In that regard, the simulation
and evaluation of the space must include aspects of the to-be
occupants, rather than solely on form generation for aesthet-
ics. Whether it is a house, industrial kitchen, or grocery store,
the building houses the space for people to perform tasks. By
SimAUD 2018 June 05-07 Delft,the Netherlands
c
2018 Society for Modeling & Simulation International (SCS)
focusing on task based evaluation, designers can improve liv-
ing and working conditions for the occupant through spatial
and geometrical design. In this paper a wide ranging literature
review introduces highly relevant and some very recent pa-
pers in biomechanics to the architectural discourse. The paper
then introduces a new method of interpreting unknown geom-
etry within CAD into a complete accessible building graph.
This is then used as the underlying mechanism in which vari-
ous human factors can be evaluated for within nearly any 3D
design program.
1.1 Human Evaluation
While often human factors in architecture are considered ther-
mal comfort levels, lighting levels, and maximal auditory
levels [14], the individual factors are less often considered.
Within academia, human factors such as the visual cone [16]
and line of sight [7] have been demonstrated as a means to
improve the design of spaces. Simulation of people through-
out the built environment has also been of interest [24, 3],
although validation of these models, and how people move in
general, is much harder to come by [1].
In implementation, both auditory and thermal levels are cal-
culated without human intervention or information on a hu-
man, and rather, are values calculated with and by the build-
ing [22, 31]. In comparison, ergonomics programs such as
Jack [2] provide industrial operations engineers the ability
to fine tune the layout and space planning of a manufactur-
ing plant based on human task and performance. Similarly,
in [27] a task based narration of agents in space is demon-
strated, however the evaluation of these tasks is not quanti-
fied.
The integration of human task based work being a metric for
design in CAD can be seen as early as the 1980’s [19]. This
work furthered into the development of Body-Motion En-
velopes, spaces required for achieving specific tasks, through
the video recording of multiple subjects attempting the
task [20]. This type of analysis has largely remained out-
side of architecture and design. One argument is that a space
can be transformed to facilitate a wide variety of tasks. How-
ever, there are numerous instances in which particular human
factors may never change, such as a disability, or a space in

182
which the purpose is unlikely to change, such as an airport,
or the interaction will always remain, such as a staircase.
As the academic fields continue to improve understandings
of how the human body moves, the knowledge found has a
great impact on the way buildings are designed, particularly
as the way architects have understood the human has directly
translated to the types of designs made [29]. In the case of
staircases, the dimensions of the stairs are defined by build-
ing codes. Through human subject studies of stair climbing,
researchers are able to understand the impact of various stair
dimensions on human movement, and as such, correlate stair
dimensions to the risk of falls and injuries [25]. Trials study-
ing the impact of cross-slope ground on human gait have
demonstrated a change in joint kinematics affecting energy
expenditure [10]. In uneven or irregular surfaces an increase
in hip and knee flexion has been seen [9]. In these exam-
ples biomechanical data has shown the built-environment has
a direct and measurable impact on the well-being of building
occupants at a physiological level.
1.2 Motivation
Although it is impossible to predict how building occupants
will interact with a space, research in areas such as Biome-
chanics, Sociology, and Neuroscience offer critical informa-
tion as to the range of possibilities. For example, research
has shown that specific decibel levels can be correlated to
specific actions of a child with Autism Spectrum Disorder
(ASD) [17]. Disabilities regarding gait have significant neg-
ative effects on life satisfaction, and are primary causes of
depression in these patients [4]. Likewise, it is dangerous for
architecture to assume medical devices are at a point of en-
abling building occupants with disabilities to successfully,and
healthfully, navigate a space, with research showing close to
85 percent of doctors dealt with patients who could not prop-
erly operate power wheelchairs because they lack the motor
skills or strength to do so [13]. With that being said, it is
also not always in the patients best interest to resort to power
wheelchairs, as their use can lead to decreased movement
and exercise, which ultimately can cause more health issues.
Therefore, conditions such as Cerebellar Ataxia (ACA), in
which people are susceptible to an increased risk of falls due
to a lower than average local stability of the trunk [5], have
profound impacts on the lives of people and their interactions
with the built environment. To assist in evaluating space for
people with this condition, a fall score is presented in 2.
Importantly, it is not only cognitive or physical disabilities
in the common understanding that can be evaluated, but also
diseases or injuries that occur purely due to aging or daily
life. Even when considering perfectly able-bodied people, re-
search has shown traditional methods for designing spaces,
for example the idea of keeping customers in a retail store
for longer periods of time, have actually had negative conse-
quences on the store from decreased consumer pleasure [12].
When considered in context to the rapid increase of online
sales for Amazon, it is no surprise that additional factors such
as accessibility, maneuverability, and shelf management [30]
were found to be key indicators for occupant happiness. Fur-
ther discussed in 1.3, past work has developed graphs to cal-
culate walking paths within design, but rarely are the graphs
refined enough to consider issues such as shelf management.
Another aspect of occupant interaction with the built envi-
ronment is reachability, especially in relation to kitchen de-
sign where a large percentage of house work is done [34].
Interestingly, research has shown that various counter heights
facilitate a reduced energy expenditure and increased comfort
when designed per task [18], yet varying counter heights are
rarely seen. Similarly to issues of counter heights, in gen-
eral poor kitchen design can cause stress on cardiovascular,
muscles, and the respiratory system [18]. Most relevant to
the research presented in this paper, research has shown cab-
inets or objects should be placed within what is referred to as
the critical reach boundary in which arm-only movement is
required [6].
1.3 Spatial Analysis
A key challenge in evaluating a design is in the interpretation
of the building geometry into a graph based system whereby
search algorithms and point locations can be used for local-
ized building information. In [15], a voxel based system is
demonstrated, with the advantage of having uniform grid ob-
jects for performance analysis. However, like most of the
research involving building analysis by graphs, the geome-
try of the building is assumed. This assumption is logical
when considering buildings designed using BIM. In [21], a
universal circulation network is introduced whereby the con-
cave and convex points of the geometries are connected to
define a building graph. To evaluate which nodes are able to
be connected an intersection function is applied to each con-
nection. While using the same approach for the graph, [11]
changes to a grid based system for alternative performance
metrics. Likewise, an even grid placement was demonstrated
in [23] with the use of geometry-to-edge intersections to cull
the edges of the graph, similar to the visibility intersections
in [21]. On a more human scale, isovists and visibility graphs
approach the problem from a starting point valid for a build-
ing occupant and continue to generate relevant information
outwards [32, 33].
Another aspect to spatial analysis, and a key component of
why evaluations are vital, is within the classification and gen-
eration of space, most recently with machine learning. Bridg-
ing these two parts, [26] demonstrates a method for using 3D
isovists to parse a scene geometry, similar to machine vision,
and then uses machine learning for classifying it. In both the
evaluation and generation side, [8] aids in rapid generation of
space plans, by automatically evaluating with custom metrics
such as number of patient beds with external window, mini-
mum distance to nurse station, maximum number of patient
bedrooms etc. useful for healthcare planners and developers.
It is this latter part, the metrics in which the space is evalu-
ated by, that has the potential to create meaningful changes
to people in a design when human factors are placed as the
priority.
During the design process a fully known building or geometry
may be difficult, especially when not using BIM. Addition-
ally, the placement of an even grid along the entire building
scene and working backwards by culling edges can present

183
inefficiencies as much of the space may not be accessible
at all. Furthermore, while the UCN is able to connect all
nodes for the shortest path [21], the resolution of the grid in
places that do not have geometry is minimal, making addi-
tional types of analysis (such as field of view from the room
center) difficult, if not impossible. Instead, this paper demon-
strates a method for generating a building graph network con-
sisting only of accessible locations for a building occupant
through an iterative raytracing approach to grow the network.
Similar to [32], a valid start node is selected by the user. The
graph is then created by iteratively checking neighbours of
each node and comparing the height at which the neighbours
exist. Further described in 2, this approach allows for a floor
surface to be made of any number of geometries while con-
sidering objects that may interfere with walkable paths.
1.4 Summary
This paper presents methods for evaluating space, particularly
within CAD programs, based on tasks performed and relevant
factors within that space. Various aspects of humans that can
be, but have traditionally not been, integrated into the design
process for evaluation were presented in the literature review
of Section 1.1. Algorithms for building a complete graph net-
work representing the movable spaces are presented in Sec-
tion 2.1, along with methods for calculating reach volumes
( 2.3) and scoring areas based on the dangers of a fall ( 2.2).
In Section 3, the software-specific implementation details of
the algorithms are shown alongside the values associated with
the designs in various configurations.
2 METHODOLOGY
When evaluating a building, in particular in the case of a
human moving throughout a space, the evaluation needs to
breakdown the model to find what areas and which ways are
possible to walk on. In the case of BIM, this is easily iden-
tifiable as this information is built alongside the geometry.
However, in many types of design software, the polygonal
or nurbs based models are created and modified by the user
without additional information embedded. For this case, a
building in which the geometry exists as un-ordered and un-
classified objects within a scene requires a graph creation sys-
tem able to iteratively create the graph with relatively little
information on the geometry. In implementation, this acts
quite similarly to autonomous robots using LIDAR to build a
scene graph as it moves along in space. This graph can then
be used for path finding and search algorithms, as well as an
equally spaced node graph used for performance metrics. In
this section, the method in which this graph can be created is
introduced. Likewise, methods for evaluating space based on
human tasks and metrics, such as reachability and dangers of
falling, are demonstrated by taking advantage of the ability to
fine tune the graph with known locations an occupant is able
to be at.
2.1 Graph Creation
Assuming a starting location that is physically possible to be
at, such as the door entrance, hallway or center of a room, an
expanding search of continuous ground can be analyzed. The
paths that a person can take are based on the edge values E
within the space (Eq 1) between nodes, denoted by r.
E={(i, j)|(ri−rj=1∨ri−rj=√2)
∧(ri−rj)·ˆz=0}(1)
This set Eis created from the calculated node set N:
N={i|i∈(j, k)|(j, k)∈E}(2)
While Nis dictated by the possible edge connections, it is
useful on its own as it defines all valid walkable surfaces, and
acts as a base for future calculations when defining regions of
a space that are impacted by other factors, such as the prox-
imity to objects.
To further illustrate the method of creation for N, Figure 1
(a) shows the parent node as valid, with 3 children invalid. In
Figure 1 (b), the relationship between what is considered a
walkable surface and none walkable surface is demonstrated
by the varying height of the surface being analyzed in com-
parison to the parent node.
Ray
Valid
20’
12’
Floor
Invalid
Parent Voi d
Valid
ab
Figure1.(a)Aparentnodeinblackissurroundedby8possiblechildren,
three(denotedbyanx,areconsideredinvalidbythecriteriashownin(b)
wherebythedifferenceinheightfromtheparentnodetothechilddemon-
stratesanimpossibleedgeconnection.
Onceavalidstartinglocationisdefined,thealgorithmcon-
tinuestoparsethesceneandbuildthecompleteaccessible
buildinggraph(Algorithm2.1).Thefunctionforbuildingthe
graphstoresallvalidnodesandedgesfound.Ateachitera-
tionofaparentnode,thesetofedgeweightsisdefinedforthe
parenttothechildren.Eachedgeisonedirectional,meaning
achildnodedoesnotautomaticallyconnecttotheparent.In
thisimplementation,theedgeweightisdefinedasthedis-
tancebetweenthetwonodes,althoughfutureworkcouldex-
pandonthis.
Algorithm2.1:BUILDGRAPH(parent,nodes,edges)
comment: Build from parent and list of valid nodes,edges
children ←GETNODES(parent)
nodes ←children
edges ←SETEDGEWEIGHT(parent, children, edges)
for node in children
do comment: Recursive Call to buildGraph
BUILDGRAPH(node, nodes, edges)

184
Within the node creation function, the algorithm checks if a
child is valid or not(Algorithm 2.2). For each node, all 8 pos-
sible child nodes are found by combinations of (i, j)of either
−1,0,or1. Note that these values represent directions, and
can be multiplied by a factor to change the spacing. Shown
as the function isWalkable, the API of the CAD program is
used to send a ray in the −zdirection and check against the
geometry in the scene for intersections. The closest intersec-
tion is used as the comparison to the parent nodes location to
define if this node is walkable.
Algorithm 2.2: GETNODES(parent, nodes)
comment: Check if parent has children not in nodes
for i←−1to 1
do ⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
for j←−1to 1
do ⎧
⎪
⎨
⎪
⎩
loc ←(parent.x +i, parent.y +j)
floor ←ISWALKABLE(loc)
if floor
then nodes ←loc
return (nodes)
PathFinding
Asthenodesrepresentawalkablesurface,andarebuiltfrom
aknownvalidlocation,thespacingofthenodescanbeused
asaquickmethodfordeterminingifaparticularpathispossi-
ble.AftercreatingtheedgesetE,Djikstra’ssearchalgorithm
canberuntofindtheshortestpathbetweentwopoints.Ifthe
spacingofthenodesissuchthatitaccountsforawheelchair,
oracomfortablebufferzone,thenodecreationwillineffect
createpathsthatcanonlyexistwiththosecriteria(FiguSF2).
abc
Figure2.(a)Anodegraphiscalculatedalonganunknownsurface.Areas
inwhichgeometry(walkableobstacles)coveredthesurfaceareshowntonot
haveanynodes.(b)Ahighresolutiongraphisshowntofindapathbetween
theflooredgeandobstaclewhereasalargergraphspacing(c)showsanot
wasnodecreatedandthesearchpathfindsadifferent(andlonger)route.
LocatingItems
Oncethenodesetisdefined,correlatinganodetoaparticu-
larobjectinthescenecanprovideopportunitiesforagranu-
larevaluation.Asshownin2.3,evaluatingkitchencabinets
forreachabilitycanbedonebyfirstassociatingn odesofthe
graphtoparticularcabinets.Thisassociationisdonebyfirst
definingaparticularo bjectwithal ocationC
,andthenfind-
ingtheclosestnodetoit(Eq3).
iC= min(N)
where i≤j⇐⇒ f(ri)≤f(rj)
f(ri)=ri−
C
(3)
In this case the objects being queried must have some infor-
mation associated with them. As shown in Figure 3, the mini-
mal amount of information required is the explicit knowledge
that a particular item is the object of interest. In practice,
the geometry can be used to find a centroid which defines the
general objects location.
1‘
2.5‘
3.5‘
1.5‘
Closest Point
Center Of Objec
t
Object
Points In Graph
Obstacle
Figure3.Visualdescriptionofhownodesareassociatedtoobjects.An
obstaclepreventsnodesfrombeingcreatednexttoanobjectofinterest.The
centroidofanobjectisfound.Thedistanceofallnodestotheparticular
centroidarecalculated.Thisprocesscanthenberepeatedforeachobjectof
interest.
2.2FallScore
AnotheruseofNfromSection2.1isindefiningareasthat
maybeconsidereddangerousforfalling.Asmentionedin
Section1.1,avarietyofpeopleranginginphysicalability
arevictimtofrequentfalling.Inordertoidentifyareasof
mostconcern,analgorithmcanbeusedtofindintersections
ofnearbyobjects.Inthesimplecase,thelevelofdangeran
objectposescanbeviewedasafunctionofthedistancea
personcouldfallbeforehittingthisobject.Thereforetheas-
sumptionthismethodtakesistheheightofapersonprojected
asahemispherearoundeachnode.
FromeachnodethenumberofoffsetsLwithintherangeof
theheightRofapersoncanbesampledby(Eq4)givenW
numberofsamples.
L=R
W−1(4)
For example, assuming an occupant height of 180cm, 6 sam-
ples would be offset at 36cm increments. At each height in-
crement the corresponding radius of intersections, a, can be
found by the equation for a spherical cap, where bis the layer
number of the current height increment (Eq 5).
O=L∗b
h=R−O
a=h(2R−h)
(5)
Using this calculated hemisphere, a function assigns values to
each node in the graph based on the number of intersections
with surrounding geometry. While not seen by the designer,
Figure 4 shows the underlying method within an example de-
sign. Likewise, Figure 5 breaks down the process for deter-
mining the point to score at each intersection.
These layer to geometry intersections are used as a way to
evaluate the impact of a fall at a nodes location. As head
trauma would be the most severe, the motivation for this scor-
ing system is based on the impact in which a fall would cause

185
ab
Figure4.(a)Avisualizationofjusttwonodesinaspacebeingscored
wherebyahemisphereisdrawnaroundthenodesand(b)acloseupofone
nodeinwhichonelayerisshownintersectingwithgeometry,visualizedby
pointsattheintersections.
onthehead.Inthisway,intersectionslowesttotheground
andfurthestawayfromthecenterhavethehighestscore,re-
latingtothefactafalltothatpositionwouldcarrythehighest
momentum.Comparatively,anintersectionclosetothecen-
ter(ie,nodelocation)andhighonthelistoflayerswouldbe
moresimilartoabumpontheheadatrelativelylowspeed.
Object Edge
Fall Radius
Object
Height
Fall Radius
l
d
/ŶƚĞƌƐĞĐƟŽŶ
++
Safe
Height
Hit
L1
L0
L2
L3
L4
L5
a
bc
Figure5.(a)Theheightofapersonisusedtodefinetheradiusofasphere,
whichissplitintoahemisphereabovetheground.Givenadesirednumber
ofsamples,slices,orlayers,(referencedasl)arecheckedagainstobjects
inthescene(b).Whenasliceintersectsageometry(c),pointsalongthe
intersectionarecheckedfortheclosestproximitytothecenterofthatslice.
ThescoreSiforeachnodei∈Nisdefinedbyequation6.
Si=
L

l=0 
g∈Gh
O+(ri+Olˆz)−pg(6)
Where riis the location of node iand pgis the closest inter-
section point of a geometry to the center of the layer. These
are summed over two components, Gand L, where Gis all
geometry in the scene that intersects with the layers and Lis
the number of layers to sample.
This function is applied to all nodes and is visualized to the
user by a normalized color bar (seen in 3.3) as the number of
objects in the scene will impact the maximum scores. While
a maximum value can still be given, comparing multiple de-
signs through the visualization becomes more difficult given
the wide ranging values. If implemented as an optimizing
feature for space planning, the number and types of geometry
being created could be controlled for, giving a more consis-
tent comparison metric.
2.3 Reach Volume
For calculating the natural reach locations from a given loca-
tion, a simple forward kinematics calculation can be used.
While this does not take into account the complex move-
ments, muscle, and tendon restraints, it does provide a gener-
alized start for analyzing maximal reach volumes. Each joint
of the arm; Shoulder, Elbow, and Wrist, are defined as ho-
mogenous transformations multiplied sequentially to deter-
mine the calculated wrist position 3
0Tas seen in Eq 7.
3
0T=1
0T×2
1T×3
2T(7)
Thereachvolumerequiresmeasurementsofthetargetarm
fromshouldertoelbow(d1),elbowtowrist(d2),andwrist
tohand(d3).Unlikethereachenvelope,thereachvolume
accountsfortheextralocationsthatelbowrotationallowsfor
acrossthechest.Similarly,theusedreachvolumepercentage
takesintoaccountthedepthacabinetcouldbetowardsaan
occupant(FigVSF6).
Valid Node
Obstacle
ĂůĐƵůĂƚĞĚWŽƐŝƟŽŶ
/ŶǀĂůŝĚEŽĚĞ
T0
T
1
T
2
T
3
T4
d
1
d
2
d
3
Figure 6. The world frame T0is transformed to the shoulder frame T1. The
reach envelope is shown by the dashed line, while the reach volume is shown
by the dots. As the obstacle prevents the person from reaching the back of
the cabinet, the ratio of valid nodes to invalid nodes shows the unused space
of that persons reach.
Each joint is defined by a sequential multiplication of transla-
tions representing the limb segment length (ie. distance from
shoulder to elbow joint) and the axis of rotation possible on
that joint.
i
jT=tx,y,z ×Rx×Ry×Rz(8)
The calculated hand position 
Pchecked against used space is
taken from the final frame 3
0T4th column as (Px,P
y,P
z).
3
0T=R
P

01(9)

186
The rotation matrices are iterated over a range of joint angles
to build the set of all reachable locations, however spacing be-
tween angles can be increased to reduce computational time.
The initial transformation 1
0Tfor the shoulder frame is deter-
mined by the direction of the closest node rCto the object
centroid 
Rand the predefined shoulder height sh.

d=rC−
R
θ=atan2(
d·ˆy, 
d·ˆx)
1
0T=t(rC+shˆz)×Rz(θ)
(10)
The final used reach volume score is the percentage of valid
nodes:
|V|
|V|+|I|×100 (11)
where Vis the set of valid nodes and Iis the set of invalid
nodes.
3 CASE STUDIES
In this section, a few example cases for using the algorithms
described in Section 2 are demonstrated. The methods are
implemented in Rhino 3D software using the Python API.
3.1 Path Accessibility
ab
Figure7.(a)Akitchenlayoutwithacutoutandspacingbetweentheisland
andsink,allowingforenoughspacetoaccomadateawheelchairwiththe
searchpathdrawnasaline.(b)Anislandtooclosetothecountertop,pre-
ventingnodesfrombeingcreatedinthatspaceandasearchpathwrapping
aroundtheisland.
Althougheachnodecouldbecheckedforintersectionsofa
wheelchairradius[28]orbufferzone[21],theuseofasparse
griddefinedbythesizen eededforawheelchairp rovidesa
quickmethodforcheckingtheaccessibilityofaspace.Fig-
ure7showsacasestudyoftwodesignswithdifferentkitchen
islandsizesandplacement.Givenaclearanceforwheelchair
accessibility,thenodesaredrawnonlyinlocationsthata
wheelchairwouldbeabletoaccess.
3.2Reach
Acommonprobleminkitchenlayoutdesignsisanunder-
utilizationofspaceduetothelackofreachability.Whileone
approachtothisanalysisistolookatthevolumeofcabinet
spaceused,thisworkusesthevolumeofreachused.Inthis
way,thefocusisontheredesignofthecabinettothehuman,
a
bc
Figure8.(a)Thetoviewofareachvolumewithfewersamplesontheelbow
iteration.A110◦anglecanbeseeninanoutervolume,withtheinnervolume
extendingfurther,representingtheelbowrotationaftertheshoulderhasbeen
fullyrotated.(b)Aperspectiveviewof(a)fromthestandpointadesigner
wouldbeusingthevisualization.(c)Areachvolumedefinedatahigher
resolutionwhilealsoshowingtheheightofthecabinetsinconflictwiththe
occupantheight.
ratherthanassumingthehumanconformstoadifferently
shapedcabinet.Inthesimpleform,thismethodprovidesa
visualizationtothedesignerofthereachvolumeaperson
wouldhaveinrelationtothecabinets,whichmayhelpinin-
stantrecognitionofsignificantdesignissues(FigVSF 8.
Alterna-tively,scoresthemselvescanbereturnedforagiven
analysis.ThedesignsinFigure8werecreatedtodemonstrate
avarietyofcabinetformsthatprovidedifferentusageratesof
thereachvolume.Figure9(a)comparedtoFigure9(b)
demonstratesa18%differenceinusedreachvolumeby
cuttingspaceforapersonintothecountertops.
ab
cd
Figure 9. (a) A typical counter and cabinet arrangement with a score of 12%.
(b)Acutoutspaceinthecountertopsallowingforapersontomovecloserto
thecabinetscoring30%.(c)Showsaslightlymoreunconventionalmethod
withcylindricalcabinets,withascoreof59%.(d)Amuchmoreradical
redesignofacabinet,althoughutilizing84%ofaperson’sreachvolume.
3.3Falling
Thiscasestudyusesakitchenlayouttorepresentthemost
dangerousareastofallinvariousdesigns.Theresultsof
thisimplementationaregiveninbothavisualizationandnu-
mericallyoutput.Inthevisualization,valueswithineach
casestudyarenormalizedtoacolorbar(FigVSF 10),while
thenumericvaluescanbeusedtodirectlycompare
variations(FigVSF 11).Thisvisualizationprovidesa
designerwiththe

187
Figure10.Fourkitchendesignswiththefallscoreevaluationapplied.The
rightcolorscaleisfromdarkblueat0(lessdangerous)todarkredat1(most
dangerous).Designs(a)and(c)havethesamespatialrequirements,withthe
sq.ftoftheislandandtablesequal.Design(b)showsaversionwithno
chairs,reducingthenumberofpossibleobjectstofallon.(d)Showsacut
outontheislandprovidingasafezoneforanoccupant.
Figure11.Histogramoftheresultingvaluesfromthefallscoredesignop-
tionsin10.Thegraphdemonstratesthatdesign(b)providesthelargestnum-
beroflowscoring(safe)zones,whiledesign(c)hasthemostareasofhighly
scored(dangerous)zones.
mostandleastdangerousareastofall,butdoesnotgivethe
fullpicturewhencomparedtootherdesigns.Forexample,
adesignermayusethisvisualizationtoeitherinformtheir
clientormodifyotherareasofthekitchenaftervisualizing
thesaferzonesassociatedwithFigure10(d)intheconcave
areaoftheisland.Additionally,Figure10(b)representsthe
dangersofmultipleobjectsandthelightbluezonearoundit
representingtheareasinwhichanoccupantcouldstillhitthe
tablefromafall.
4DISCUSSION
Therearemanyareasforexpansionandimprovementinthese
evaluationmethods.Inrelationtothefallscore,therewould
besignificantimprovementsbyconductingfallt rialstofind
thelikelihoodofacertainlocationbeinghit.Alongsidehu-
mantrials,physicssimulationscouldbeintegratedinCAD
programstomodelthetestedbehaviour.Oneproblemwith
thefallscoremethoddemonstratedhereisthelackofin-
tegratingorderedobjects,suchthatawallclosertotheoc-
cupantwouldnotberecognizedasblockingtheoccupant
fromfallingonanobjectbehindit.Withinreachevalua-
tions, past work has considered how much of a cabinet an
occupant could reach, while this looks at the utilization of the
occupants reach. Future work could incorporate both of these
metrics, as well as finer detailed information such as approach
angle for grasping. In the case of path search, a great deal
of research has been done with walking distances, however,
evaluating energy expenditure requires human subject inter-
vention. Ideally, understanding the path travelled, given the
same distance, requires different energy for carrying heavy
bags of groceries from the super market than an envelope at
the post office. Finally, the goal of this work is to demonstrate
both the method for a complete accessible building graph and
evaluation methods of human factors beyond those that relate
to thermal or auditory comfort that can be integrated into the
automated design workflows being researched presently.
ACKNOWLEDGMENTS
The authors thank Lynn Bongiovanni for her valuable dis-
cussions in kitchen design and work on early versions of the
project. The authors also thank Christopher Gornall for his
insightful literature review and early work on identifying ele-
ments of Autism Spectrum Disorder that can be quantified in
relation to building evaluation.
REFERENCES
1. Berry, J., and Park, K. A passive system for quantifying
indoor space utilization. In ACADIA 2017, Association
for Computer Aided Design in Architecture (ACADIA)
(2017), 138–145.
2. Blanchonette, P. Jack human modelling tool: A review.
Tech. rep., 2010.
3. Breslav, S., Goldstein, R., Tessier, A., and Khan, A.
Towards visualization of simulated occupants and their
interactions with buildings at multiple time scales. In
Proceedings of the Symposium on Simulation for
Architecture & Urban Design, Society for Computer
Simulation International (2014), 5.
4. Broe, G. A., Jorm, A., Creasey, H., Grayson, D.,
Edelbrock, D., Waite, L. M., Bennett, H., Cullen, J. S.,
and Casey, B. Impact of chronic systemic and
neurological disorders on disability, depression and life
satisfaction. International journal of geriatric psychiatry
13, 10 (1998), 667–673.
5. Chini, G., Ranavolo, A., Draicchio, F., Casali, C., Conte,
C., Martino, G., Leonardi, L., Padua, L., Coppola, G.,
Pierelli, F., et al. Local stability of the trunk in patients
with degenerative cerebellar ataxia during walking. The
Cerebellum 16, 1 (2017), 26–33.
6. Choi, H. J., and Mark, L. S. Scaling affordances for
human reach actions. Human movement science 23,6
(2004), 785–806.
7. Ciftcioglu, O., and Bittermann, M. S. Fusion of
perceptions in architectural design.
8. Das, S., Day, C., Hauck, J., Haymaker, J., and Davis, D.
Space plan generator: Rapid generationn & evaluation
of floor plan design options to inform decision making.

188
In ACADIA 2016, Association for Computer Aided
Design in Architecture (ACADIA) (2016), 106–115.
9. Dixon, P., Sch ¨
utte, K., Vanwanseele, B., Jacobs, J.,
Dennerlein, J., and Schiffman, J. Gait adaptations of
older adults on an uneven brick surface can be predicted
by age-related physiological changes in strength. Gait &
posture 61 (2018), 257–262.
10. Dixon, P. C., and Pearsall, D. J. Gait dynamics on a
cross-slope walking surface. Journal of applied
biomechanics 26, 1 (2010), 17–25.
11. Doherty, B., Rumery, D., Barnes, B., and Zhou, B. A
spatial query & analysis tool for architects. In
Proceedings of the 2012 Symposium on Simulation for
Architecture and Urban Design, Society for Computer
Simulation International (2012), 4.
12. Donovan, R. J., Rossiter, J. R., Marcoolyn, G., and
Nesdale, A. Store atmosphere and purchasing behavior.
Journal of Retailing 70, 3 (1994), 283 – 294.
13. Fehr, L., Langbein, W. E., and Skaar, S. B. Adequacy of
power wheelchair control interfaces for persons with
severe disabilities: A clinical survey. Journal of
rehabilitation research and development 37, 3 (2000),
353.
14. Frontczak, M., and Wargocki, P. Literature survey on
how different factors influence human comfort in indoor
environments. Building and Environment 46, 4 (2011),
922–937.
15. Goldstein, R., Breslav, S., and Khan, A. Towards
voxel-based algorithms for building performance
simulation. In Proceedings of the IBPSA-Canada eSim
Conference (2014).
16. Hudson, R., and Westlake, M. Simulating human visual
experience in stadiums. In Proceedings of the
Symposium on Simulation for Architecture & Urban
Design, Society for Computer Simulation International
(2015), 164–171.
17. Kanakri, S. M. F. The impact of acoustical
environmental design on children with autism. Texas
A&M University, 2014.
18. Kishtwaria, J., Mathur, P., and Rana, A. Ergonomic
evaluation of kitchen work with reference to space
designing. Journal of Human Ecology 21, 1 (2007),
43–46.
19. Lantrip, D. B. Isokin: A quantitative model of the
kinesthetic aspects of spatial habitability. In Proceedings
of the Human Factors Society Annual Meeting, vol. 30,
SAGE Publications Sage CA: Los Angeles, CA (1986),
33–37.
20. Lantrip, D. B. Environmental constraint of human
movement: A new computer-aided approach. In
Proceedings of the Human Factors and Ergonomics
Society Annual Meeting, vol. 37, SAGE Publications
Sage CA: Los Angeles, CA (1993), 1043–1043.
21. Lee, J.-k., Eastman, C. M., Lee, J., Kannala, M., and
Jeong, Y.-s. Computing walking distances within
buildings using the universal circulation network.
Environment and Planning B: Planning and Design 37,
4 (2010), 628–645.
22. Marsh, A. Ecotect and energyplus. Building Energy
Simulation User News 24, 6 (2003), 2–3.
23. Nagy, D., Villaggi, L., Stoddart, J., and Benjamin, D.
The buzz metric: A graph-based method for quantifying
productive congestion in generative space planning for
architecture. Technology— Architecture+ Design 1,2
(2017), 186–195.
24. Narahara, T. The Space Re-Actor: walking a synthetic
man through architectural space. PhD thesis,
Massachusetts Institute of Technology, 2007.
25. Pauls, J. L., Fruin, J. J., and Zupan, J. M. Minimum stair
width for evacuation, overtaking movement and
counterflowtechnical bases and suggestions for the past,
present and future. In Pedestrian and evacuation
dynamics 2005. Springer, 2007, 57–69.
26. Peng, W., Zhang, F., and Nagakura, T. Machines
perception of space: Employing 3d isovist methods and
a convolutional neural network in architectural space
classification. 474–481.
27. Schaumann, D., Kalay, Y. E., Hong, S. W., and Simeone,
D. Simulating human behavior in not-yet built
environments by means of event-based narratives. In
Proceedings of the Symposium on Simulation for
Architecture & Urban Design, Society for Computer
Simulation International (2015), 5–12.
28. Schwartz, M. Collaborative and Human Based
Performance Analysis. In eCAADe: Models of
Computation - Human Factors, vol. 2, Faculty of
Architecture, Delft University of Technology (Delft,
2013), 365–374.
29. Schwartz, M. From human inspired design to human
based design. In Morphological Analysis of Cultural
DNA. Springer, 2017, 3–13.
30. Teller, C., Gittenberger, E., and Schnedlitz, P. Cognitive
age and grocery-store patronage by elderly shoppers.
Journal of Marketing Management 29, 3-4 (2013),
317–337.
31. Trnsys, A. Transient system simulation program.
University of Wisconsin (2000).
32. Turner, A., Doxa, M., O’sullivan, D., and Penn, A. From
isovists to visibility graphs: a methodology for the
analysis of architectural space. Environment and
Planning B: Planning and design 28, 1 (2001), 103–121.
33. Turner, A., and Penn, A. Evolving direct perception
models of human behavior in building systems. In
Pedestrian and Evacuation Dynamics 2005. Springer,
2007, 411–422.
34. Yazıcıo ˘
glu, D. A., and Kano˘
glu, A. A systematic
approach for increasing the success of kitchen interior
design within the context of spatial user requirements.
Advances in Social Sciences Research Journal 3,1
(2016).