Automation in Construction: Mathew Schwartz

Automation in Construction 127 (2021) 103557
Contents lists available at ScienceDirect
Automation in Construction
journal homepage: www.elsevier.com/locate/autcon
Human centric accessibility graph for environment analysis

Mathew Schwartz
New Jersey Institute of Technology, Newark, NJ, USA
ARTICLE INFO ABSTRACT

Keywords: Understanding design decisions in relation to the future occupants of a building is a crucial part of good
Accessibility design. However, limitations in tools and expertise hinder meaningful human-centric decisions during the
Graph design process. In this paper, a novel Spatial Human Accessibility graph for Planning and Environment Analysis
Spatial analysis
(SHAPE) is introduced that brings together the technical challenges of discrete representations of digital
Walkability
models, with human-based metrics for evaluating the environment. SHAPE: does not need labeled geometry
Human factors
Computation
as input, works with multi-level buildings, captures surface variations (e.g., slopes in a terrain), and can be
used with existing graph theory (e.g., gravity, centrality) techniques. SHAPE uses ray-casting to perform a
search, generating a dense graph of all accessible locations within the environment and storing the type of
travel required in a graph (e.g., up a slope, down a step). The ability to simultaneously evaluate and plan
paths from multiple human factors is shown to work on digital models across room, building, and topography
scales. The results enable designers and planners to evaluate options of the built environment in new ways,
and at higher fidelity, that will lead to more human-friendly and accessible environments.
1. Introduction the accessibility of a space. While walkability as a term has various defi-
nitions [2,3], it is often considered to be an aggregate of transportation,
Understanding design decisions in relation to the future occupants policies, mobility, and locations of points of interest [4,5]. This paper
of a building is a crucial part of good design. However, technical uses the term walkability as the ability to walk on/over, reflective of
limitations in tools and the need for an extensive understanding of the Environments Conditions→Traversable theme illustrated in [3]. The
human physiology prevent the ubiquity of meaningful human-centric term accessibility is used in this paper as a person’s ability to access
decisions during the design process. While the topic of human move- a space—from a human factors perspective. The various conditions
ment has been discussed in the literature at length, there is no method surrounding this ability and the associated costs it has on a person, are
that approaches the construction of the graph-representation of a digital vital considerations for human comfort. This is not to say that existing
model with a physiological human factors perspective.
techniques of spatial analysis are not valid. Rather, this paper argues
When considering an environment designed for people, accessibility
for additional human-centric data to be used, and proposes a method
is one of the most fundamental criteria for evaluating the space’s
for doing so.
performance. While recent work in computer science and architecture
While there are past works that evaluate the built environment
have shown an increase in the use of data and quantified metrics
for optimization or evaluation, there has been only a few works that with regard to human-based mobility such as slope and pavement
truly consider human factors of an occupant at the physiological level. quality [2], the apparatus of the evaluation is often survey-based,
Although building design usually incorporates human factors, the types checking if a space meets a given criteria that has been generalized
of calculations used are often limited in scope. For example, the term or predefined. An important component of analyzing slope in an en-
human comfort is often used in architecture and urban design as a vironment that is often lacking in survey data is the question: ‘‘why
reference to the thermal, visual, and acoustic relationships of the envi- is this slope bad?’’. This disconnect can happen when the metric being
ronment and occupant [1]. To improve analysis, and subsequently, the used to analyze an environment relies on the comparison to building
performance of a building for its occupants, we must be able to stream- code and existing guidelines that create binary and abstracted measures
line the evaluation of a design, taking into account the complexities of (e.g., [7,8]). Additionally, while automated code checking based on
human experience in aggregate. existing guidelines is valuable for assessing workflows and minimizing
Accessibility is not a binary value, and the difficulty associated with errors and time [9], the power of computation allows for a more
moving from one point to another must be included when considering complex relationship between accessibility and regulations.
E-mail addresses: cadop@njit.edu, cadop@umich.edu.

URL: http://www.cadop.info.
https://doi.org/10.1016/j.autcon.2021.103557
Received 21 September 2019; Received in revised form 23 December 2020; Accepted 15 January 2021
Available online 23 April 2021
0926-5805/© 2021 Published by Elsevier B.V.
M. Schwartz Automation in Construction 127 (2021) 103557
The two core contributions of the paper to the field are summarized
as follows:
1. A novel ray-casting based method for creating a grid-based

weighted directed graph that enables early-stage feedback in the
design process with respect to human experience and accessibil-
ity.
2. An in-depth review of the current literature that categorizes and
illuminates factors that should be included in path and circu-
lation analysis both indoor and outdoor for ensuring human-
centric built environments.
The paper is organized as follows; Section 2 presents a review of

literature on human factors relating to accessibility and motivation for
such an approach, Section 3 provides an overview of the types of data
Fig. 1. The three environmental scales of City, Building, and Object, are all integrated.
structures and algorithms for evaluating 3D models, Section 4 intro-
Path analysis and accessibility is from one point to another, and includes numerous
environmental conditions.
duces a novel algorithm for parsing a building into a graph, Section 5
Source: Adapted from [6]. demonstrates the presented work in a design context with various scales
and environment types, and Section 6 discusses limitations and future
work.
Unless ultra-strict laws and regulations are enacted for explicitly
2. Human factors
limiting what is allowed to be an accessible path, it is not feasible
for a designer to consider these conditions. Furthermore, this extreme
Evaluating the accessibility of a space concerns not only Euclidean
restriction-based approach limits the design as well as opportunities to
distance but also factors such as energy expenditure, speed, slips, trips,
improve accessibility through innovation. Computation of human fac-
and falls. Likewise, analyzing an environment for human factors and
tors combined with automatic code checking provides an ideal platform
comfort must include the tasks and movements to be performed in any
for – performative rather than prescriptive – regulation. In order to given location (e.g., cutting vegetables or opening a cabinet). In this
achieve this goal, we must first solve the challenge of finding acces- section, the basis for a weighted digraph built from an accessibility
sible locations, and subsequently, incorporating more than distances standpoint is introduced.
between the locations to properly represent the occupants experience. While some laws and regulations exist around the type of floor sur-
In prior work, human-based accessibility analysis has been per- face of a building or urban environment such as the American Disabili-
formed after the generation of a navigable surface. For example, a ties Act and its accessibility guidelines [10] for built environments—
navigation mesh is first defined; then, a dense graph is generated within algorithmic methods for quantifying the accessibility of these envi-
it, and a calculation is performed on accessibility factors thereafter; ronments are lacking. However, a large body of research exists that
and finally, the graph is culled to the meaningful components. In this quantifies human biomechanics in relation to the built environment.
paper, a novel approach is presented for the first step – defining the By developing a method for interpreting a 3D model based on the way
graph – with a focus on human accessibility, enabling a scalable and in which a human moves within a space, the literature based on human
information-rich dataset. The graph is constructed by rules dictated factors can be applied and used as a more informative metric than
from accessibility related factors directly from the physiology and distance alone.
biomechanics literature. The algorithm is scalable from the object to From a biomechanical perspective, mobility within the built envi-
urban scale (Fig. 1) and can be used in multiple other analyses (from ronment is impacted by surface types and geometry against physiolog-
light intensity to fall probabilities) to inform the cost of accessibility. ical factors (e.g., age, weight, and disability). In Table 1, six types of
The approach allows for early design-stage feedback that can not only surfaces are identified with a description and relevant reference to the
improve designs but also allows for quick iterations and reduced costs biomechanics literature. These surface types apply to both the outdoor
(Fig. 2). and indoor environments.
Fig. 2. A multi-level building example of using the AG. Steps shown are (1) A starting model is given in the top left (2) the graph is generated finding accessible locations (3)
a viewshed-type analysis is performed to find locations with maximal space (4) path planning start and end locations determined (5) generate paths with various criteria (6)
reflection on paths generated by finding use cases for a given fire stairwell vs. the open staircase.
2
Table 1
Surface types and their significance in human mobility, with corresponding references. Surface material (e.g., sand or concrete) has an impact in all the examples; however the
references focus on angles and variation.
Surface type Description and example Reference
Flat Leveled and flat floor with little to no variation across the surface. [11]
Ramp Slope is along the progression axis. A person walks in the same direction as the incline. The balance is front to back. [12,13]
Cross-slope Slope is perpendicular to the progression axis (Fig. 3). A person walks perpendicular to the incline, forcing the body to adjust [14–16]
side to side.
Staggered Stairs and other flat surfaces that are offset and have a change in level, such as bricks that are protruding directly up in [17–20]
various heights (but not at an angle) or the offset of a curb to the sidewalk.
Uneven Variations in curvature of the ground surface (Fig. 5), such as in most natural topography, especially mountains, and some [21,22]
poured concrete curved surfaces in modern design.
Uneven staggered Faceted surfaces that are both angled and protruding, such as bricks in an old walkway with shifted ground underneath. [23]
and energy up to 60% [29]. The inclusion of this data for the analysis
of travel times to amenities in the urban environment (walkability) or
within a building would drastically change the results.
The impact of a ramp on accessibility is not a simple relationship
that every designer can instinctively work with. The increase in slope
angle of a ramp is not linearly correlated to the energy increase it
takes for a human to move up it [12,13]. As seen in Fig. 4, the energy
expenditure associated with the slope decreases in a slightly negative
gradient. As the gradient negatively increases, the energy expenditure
increases as well. Walking on a cross-slope (Fig. 3) of 6◦ has shown
a threefold increase in mediolateral ground reaction force, which is an
artifact of adaptations in movement to reduce falling [14]. The impact
Fig. 3. Cross-slope (left) is when the direction of movement is perpendicular to the
of a cross-slope for wheelchair propulsion is profound as well, with a
direction of the slope. Cross-slope forces one side to have a shorter stance than the other
when compared to a ramp (right) in which both sides equally vary during movement. 2% cross slope increasing energy by over 30% [30]. [19] studied a
staggered block surface similar to bricks with a vertical variation up
to 2.5 cm and observed a 28% increase in energy expenditure.
In a study of adults under 60 years old, it was found that traumatic
brain injuries from falls occurred most often in ground level falls (36%),
with the second most common being falls down stairs1 at 27% [31].
Stairs, especially in the urban environment, can significantly range in
height, depth, and inclination. Accounting for not only the existence of
a step or slope, but also the slope of steps, is a critical part of analyzing
an occupant’s path. Energy requirements have been shown to increase
by 67% from a 24◦ to a 42◦ incline. Moreover, the gait pattern is
notably not a linear transition between walking and stair climbing, and
it likely switches between the two at a certain threshold [32]. Crutch
ambulation is another example of the complexities in accessibility
analysis. Studies have shown occupants using crutches on stairs had
Fig. 4. Energy cost of walking at various gradients as defined by the polynomial a twofold increase in energy expenditure, while stair usage was more
function in [13] and shown in Definition 9. efficient than ramps considering energy cost per vertical rise [29]. The
caveat to this efficiency is when the ramp is less than 5% it becomes
the more efficient means of travel.
At the urban scale, the idea of a single brick out of place can seem While regulations exist for many stair dimensions, there is a need for
irrelevant. However, these small details and aspects of the environment a more cohesive analysis measurement for human mobility that takes
can significantly affect occupants. Variation in brick height of a walk- into account a combination of environmental and spatial factors and
way can have a direct and meaningful impact on the ability for people conditions. This is evidenced by [33], which demonstrated a decrease
to comfortably and safely move [23]. Irregular surfaces (e.g., gravel) in ambient lighting from 300 lx to 3 lx (limit of civil twilight) led to
require significant adaptations in gait to reduce tripping, at the cost of young adults increasing foot clearance while almost all older adults did
speed, step length, and cadence [24]. Furthermore, wheelchair propul- not. These small deviations (or lack there of) in human movement can
sion energy costs increase significantly with surface type [25]. The lead to serious consequences (i.e., falling).
net locomotive energy cost for a wheelchair on carpet compared to a It is important to note that the discussion around energy expendi-
tiled surface can increase by more than 70% [26]. Surface resistance ture should not be limited to the simple issue of calories burned. In
changes not only the energy required but also the speed and rate each case identified (e.g., stair negotiation or cross-slope wheelchair
of movement (such as a nearly 50% reduction in average velocity propulsion), an increase in energy expenditure can mean increased
from tile to carpet) [27]. In [28], the authors explicitly mentioned loads on joints and muscles, thus leading to injuries. When considering
the importance of floor type when designing a hospital owing to their surface variation and cross-slope movements, a simplistic model of a
findings in energy expenditure across concrete and carpeted floors. line connecting spaces within the built environment is not sufficient.
In [29], the authors have made a similar point about the importance A graph-based approach in which the directed edges can be weighed
for designers to know this impact when designing homes. based on a perpendicular edge (i.e., the cross slope) must be used.
When discussing disabilities and mobility an occupant using a Calculating the complexities of human mobility is vital for accurate
wheelchair is often at the forefront of the discussion. It is important to circulation and time based analysis of paths. When considering mobility
remember the wide range of human mobility and assistive devices. For
1
instance, canes and crutches may cause a reduction in mobility speed For an extensive review of falling and stair negotiation, see [20].
3
Fig. 6. Representation of the environments into a grid of nodes. Illustration showing

the geometry, nodes on the geometry for visualization, and the points representing the
environment used for analysis in computation.
Fig. 5. Example of real-world case for varying surface heights. Suwon, South Korea,
(Finished Construction in 2015).
3.1.1. Surface definition

In games and character animation, the accessible space of an agent
with a wheelchair, the option to climb steps is clearly non-existent.
is generated in various ways from a given 3D model. The surface of
However, with a cane or crutches, the question of energy expenditure
space designated to an agent’s movement is often referred to as the
and time become directly linked to the geometry of the environment.
Navigation Mesh [39]. For one, a geometry object can simply be labeled
Discussing the impact of stairs on crutch ambulation, [29] mentions
as a traversable surface. In this case, triangulation by subdivision, or
specifically the situation of transportation terminals requiring a 10 m
a uniform grid can be created, to define links to move across the
ascent in 200 m horizontal distance requiring 29% more energy and
environment [40]. Two popular approaches to the generation of a
45% more time than a 5% ramp. When considering locations of the navigable space are (1) reconstruction through voxelization [41] and
start and end of the steps, as well as the impact different ramp and step (2) mesh vertex calculations [42,43]. Other methods such as multiple
gradients would have of these endpoints, the only realistic solution for horizontal slices with depth-maps [44] and prismatic subdivisions [45]
ensuring human well-being and health is at the forefront of a design is have been shown, with a target of fast agent-based simulation at the
the automated evaluation and feedback to a designer. expense of pre-processing.
Notably, software such as Unity [46] implements an automated
3. Related work and motivation generator using voxelization for a NavMesh to define surfaces on which
a virtual agent can move. However, since architecture models might
have thousands of objects and millions of polygons detailing the space
Evaluating a design is a long studied and complex topic. To illustrate
(e.g., molding sand decorations), especially objects not related to ac-
the contribution of the proposed method, the following sections break
cessibility (e.g., HVAC systems), this approach is inefficient. While
down the problem into two main areas; the data structures that are
ignoring these unrelated objects is possible, this means the geometry
used to define where an occupant could be in an environment (Sec-
is already labeled and the walkable surfaces are known. However,
tion 3.1 and the methods for evaluating an environment (Section 3.2).
accessibility related issues are still not known even when intended
As the method for creating an Accessibility Graph presented in this
walkable surfaces are defined (e.g., a ramp is a walkable surface but
paper brings together these two, often disconnected components, the
the slope is too steep).
literature presented is highly varied in disciplines. Alternative methods considered to be true representations thus use
vertices of the mesh itself. For generic path-planning of a character
3.1. Navigation data structures the parsing of the environment with little information attributed to
the graph is often suitable. For example, the undirected graph created
In robotics, path-planning of an environment for both humanoid and from an Explicit Corridor Map [47] or the Implicit Corridor Map [48]
wheeled robots (e.g., autonomous vehicles) is often done in unknown inherently embedded relative obstacle distance information enabling
environments such as the DARPA Robotics Competition [34]. In games, the generation of more realistic paths for an agent to travel along.
human players and autonomous agents must move around a space with Improvements on surface-based approaches annotate the environment
through the identification of gaps to produce a continuous navigation
directions and paths that are feasible to their goal. Within Architecture,
mesh of the entire walkable region [49].
circulation of people moving in the environment and calculating the
The focus in these cases is the ability to generate a visually realistic
times it takes to reach various locations are considered. While each
path for an agent to follow in a given environment. In design evalua-
of these fields, namely robotics, games, and architecture, is inter-
tion, the generation of an agent path for visual realism is less important
ested in the paths taken within a space, there are also multiple other
than the quantification of the space in which that agent would move.
considerations that impact the approach and use of various techniques.
It is therefore important to maintain the broad knowledge of where
The techniques for finding the navigable areas are most prominently in a model an occupant could physically access, while enabling a data
different in whether an environmental model exists or not. In robotics, structure that contains location-specific information.
the environment is often unknown and a map must be continuously
built for path-planning [35,36]. A common method is to create a height 3.1.2. Graph and map representation
map – a grid of cells labeled as a floor or obstacle depending on the A common approach to defining paths and points of interest for
height difference between them [37]. Similar to the robotics problem of spatial analysis involves explicitly defining a network of nodes rather
reconstructing real-world environments, in urban analysis a Digital El- than the construction of a polygon or surface (as seen in a Navigation
evation Map (DEM) is created through LIDAR data, identifying areas of Mesh). While the polygons of a navigation mesh are able to describe a
roughness using gradient calculations [38]. Beyond these techniques for graph, a dense graph representation of an environment (Fig. 6) has an
on-the-fly generation, either a polygonal bounds defining a traversable important difference in usage than general navigation meshes: nodes
surface is created, or a discretization is performed that represents the of the graph are used for additional analyses of the environment. In
space as a graph (nodes and edges). commonly used analysis methods of the built environment such as
4
lighting, the calculations are reduced to a grid or specific points on the 3.2.1. Agent based analysis
surfaces within the environment (e.g., [50]). This discrete analysis is A common approach to evaluating the way occupants would utilize
particularly useful when considering maximum or minimum values and a given design is to virtually recreate the individual occupants, referred
when it is not easy to know which points are important or of interest to as agents. Agents are assigned some tasks and utilize the navigation
ahead of time (e.g., finding the highest intensity of light in a room data structures described in Section 3.1 to provide feedback on how
for glare analysis). Therefore, a dense representation should be created those agents experienced the space. Likewise, the collective movement
from the start. of the agents based on those data structures can evaluate the efficacy of
Directed graphs are vital to spatial analysis: the view while going the design. Most similar to game-based evaluation, [67] implemented
in a straight line from point A to B is 180◦ from B to A. A visibility a responsive narrative in which a character moves and interacts with
graph [51] constructs a graph from the direct connections of all vertices the environment, offering subjective feedback through visualization.
Through multiple iterations of agents moving in the environment,
in a set, allowing for optimal shortest path planning; however, the
statistics and data are extracted about how the agent performed and
connection to visibility from a human perspective is lacking. This
experienced the space. Influential work for evaluating an architectural
method was used by [52] to connect a generated grid of nodes along
space was in the application of egress [68,69]. In simulating the proba-
a floor surface which had to be pre-defined by the user. Grid-based
bilistic movement of a crowd during evacuation, social and behavioral
graphs have also been constructed in a more localized manner, such
factors have significant impact on the accuracy [70]. For egress, the
that adjacent nodes (or cells) are the only points of contact [53,54].
value for analysis is relatively straightforward, i.e., length of time for
Similarly, [54] used edge culling by checking the intersections of nodes
occupants to exit. When deriving useful data on the complex experience
from an even grid on the floor to object geometry. Another approach of an occupant, the inputs to the simulation greatly increase.
defines a Nav-graph using a convex hull around vertices of a 3D ob- In addition, in a narrative-based model for multi-agent simula-
ject projected into a two-dimensional (2D) plane [55]. In particular tion, [71] created a simulation around spaces, actors, activities, and
for path-planning, a dense graph can be created first with weights narratives. The data extracted is of walking paths, density, length of
associated to locations, then construct a hybrid graph after [39,56]. stay, and social interactions. As even the same person does not act
Often in a design context, minimal, rather than dense graphs are identically on multiple days, numerous iterations of agents must be
constructed. Sometimes this is due to the input data, such as lines and simulated to demonstrate a convergence towards accurate data. The
intersecting points for road analysis [57], or the ability to use BIM challenge in this approach is the efficacy of properly representing
data to find doorways [58,59] to calculate the shortest paths between aspects such as the social interaction between agents. In true human
spaces. When constructing the graph representation of a building, interactions, aspects as small as eye contact can change the social
labeled data (e.g., BIM, GIS, IFC) is often used, rather than the geometry dynamic and experience [72]. Through deep learning, [73] developed
in isolation [58–60]. Others have used labeled models, but still provide a model to control a virtual biomechanical human to the point of
a fallback of surface normals to aid in a tessellation process for defining photoreceptor responses that inform head directions and gaze. While
navigable networks [61]. agent based simulation addresses an important aspect of time, the
While games can rely on reducing model complexity with visual difficulty in modeling true social interactions suggests, at least for the
tricks, Architectural models have complex geometries that represent time being, that more concrete metrics in the environment independent
the real infrastructure and that include details often irrelevant to of social factors must be included in spatial evaluation. For example, as
human perception or accessibility [62]. Ray-casting, as implemented mentioned previously, the curvature of a surface directly affects human
for the Spatial Human Accessibility graph for Planning and Environment occupants as they traverse it, regardless of the point in time.
analysis (SHAPE), in large and complex scenes can provide an efficient
3.2.2. Graph based analysis
alternative; for example, complex sculptures, islands, or other scene
In a more discrete approach than agent-based simulation, graph
assets that are visible, but not accessible, are quickly excluded from
techniques work towards integrating known numeric quantities about
calculations (e.g., through a bounding volume hierarchy).
a space, or attributes of locations in the space, to a graph. While in
Discussed in Section 2, slope and surface variation is vital infor-
Section 3.1.2 the specific methods of creating a graph were compared
mation to human mobility, yet methods for considering slope with a with the current work, this section will discuss the evaluation methods
graph are limited. An easiest path for space syntax was explored in [63] and criteria for analyzing a graph in design. Network analysis in the
using an abstracted environment representation based on lines (taken built environment, in its simplest form, assesses relationships between
from street segments) with a slope defined by the two endpoints. An various locations. Such spatially relevant networks consisting of nodes
indoor topology for human pathfinding which incorporates stairs and and edges have been demonstrated in BIM [58,74], but are difficult to
ramps was introduced by [59]. With a focus on the computational construct with unlabeled data. By using intersections and properties of
side, research on natural path generation considering slope and terrain the environment (3D model), networked relationships of a space can
variation has been demonstrated in [64,65]. Another technique used be constructed. Early work in automatic evaluation was in the use of
to incorporate 3D geometry is by voxelizing the environment, although polygons representing a field of view (ISOVISTS) [75]. The intersecting
these box representations make it difficult to incorporate fine-grained polygons can then be reduced to numeric values of visibility intensity
analysis such as slope and varying surfaces [66]. While these works in for a given location, which was later modified to a graph-specific
part demonstrate a use of slope to determine routes, the SHAPE method data structure [52]. More recently, [76] implemented an all-to-all
is the only one able to combine the dense graph representations without connection of nodes to construct a 3D visibility graph for space syntax
labeled geometry and include multiple factors in the graph. analysis. These analysis methods are all possible with the SHAPE, while
the 3D visibility graph can be created with human-accessible locations
by offsetting the SHAPE vertically in some discrete increments.
3.2. Common evaluation methods
While network and circulation analysis often uses the shortest path
algorithm, the physically shortest path is not necessarily a good metric
Quantifying human experiences in the built environment can be of the way in which people will move within the environment nor the
done by either simulating a virtual human (agent-based simulation) or experiences an occupant would have. With metrics such as walkability
evaluating spatial information represented as a graph (e.g., space syn- (referred to in Section 1) commonly using the path length to represent
tax, network analysis). These two options are compared and discussed a unit of time, [63] reduced multiple metrics to a single value of time
in the following sections. of travel for analysis. However, the relationship between speed, energy,
5
metric, such as visibility, information about the access to a node is

lost. This is exemplified in the problem of an environment location
not being accessible at all, while still calculating the values associated
with those locations. The approach introduced in this paper considers
the method of movement from one node to the next, allowing for a
quick end to the computation when starting from a poorly accessible
location.2 Combining the movement from one node to another with the
node based analysis, numerous considerations for generating a likely
or optimal path can be used. The factors of one directed edge cost
can be weighted against others, as well as any attribute associated
with a node. Furthermore, the movement type can be used in path
planning and analysis, where the cost of a path for a search algorithm
(e.g., Dijkstra [80]) is built on weights such as slope.
Fig. 7. The breadth-first process of defining node locations. The numbers correspond
to the order the node was checked. Two box objects are defined as obstacles while the The method presented expands the ray-casting based breadth-first
surface itself is a free form shape in 3D. Further examples are given in Appendix A search to find accessible locations of an environment given an initial
and Supplemental Materials. starting position as described in [81]. In this section, the method of the
graph generation is explained followed by example implementations of
both path planning and path analysis.
and human ability is well researched in the literature provided in Sec-
tion 2 and reducing to a time-based metric is often not representative.
In the case of accessibility, the physically shortest path may not be 4.1. Graph components
the physically least demanding nor the one an occupant may take.
The types of data used in graph-based analysis of paths are wide and 4.1.1. Graph definition
varying, such as calculating the number of turns along an occupant To define the graph of an environment, a starting point 𝜏 ∈ R3
path [6,63,77,78]. Similarly, the types of users have a profound impact within the CAD environment is defined from which a breadth-first
on the importance of different factors when considering accessibility, process begins. The graph 𝐷 ∶= (𝑉 , 𝐸, 𝑊 ) is a weighted directed graph
such as considering direct connectivity and spatial depth for elderly (digraph) composed of the set of vertices 𝑉 , edges 𝐸, and a set of weight
housing [78]. vectors 𝑊 . The set of vertices is defined as:
One challenge in a sparse network can be seen in [6], where an
Indoor Walkability (IW) factor was used to quantify paths based on Definition 1. 𝑉 (𝐷) ∶= {𝑣1 , … , 𝑣𝑛 }.
various weights. As the graph used in path generation consists of
sparse nodes connecting maximal spaces, the weighted values are path- The edge set of 𝐷 is defined as:
based, rather than information on what is along the path. The authors
proposed future work, including object-level refinement. However, it Definition 2. 𝐸(𝐷) ∶= {𝑒1 , … , 𝑒𝑘 ∣ 𝑒𝑖𝑗 = (𝑣𝑖 , 𝑣𝑗 ) ∧ 𝑣𝑖 , 𝑣𝑗 ∈ 𝑉 (𝐷)}.
is unclear how the sparse network of locations used in the UCN [58]
would function in this context. For example, while a table has geometry The last basic component of the digraph is the set of weight vectors:
that can be included in the network, overhead lighting casting a bright
spot in the middle of a room would be ignored. While the motivation Definition 3. 𝑊 (𝐷) ∶= {𝐰𝟏 , … , 𝐰𝐤 }.
of [6] is similar to the SHAPE, a key difference is the scalability and
limitations related to both requiring labeled data and limited nodes Where 𝐰 is a vector of weights with a mapping to edges such that
throughout the environment. 𝛹 ∶ 𝑒𝑖𝑗 → 𝐰𝐢𝐣 . The weight vector 𝐰𝐢𝐣 contains the different costs
While these past studies have varying resolution to their graph associated with the movement on an edge. A subscript is used to denote
structures, they share a common theme, emphasizing the utility of a specific weight in the vector, such that the weight corresponding to
graph theory in environmental analysis. However, the focus in the distance is denoted with subscript 𝑑, such that 𝛹𝑑 (𝑒𝑖𝑗 ) = 𝚍𝚒𝚜𝚝(𝑣𝑖 , 𝑣𝑗 ).
past works of space syntax and network analysis has been primarily To map the graph to a spatially relevant set of locations, the set
on the relationship between the environment and building rules & of vertices have an intrinsic mapping to 3-dimensional points in the
regulations, while the work presented here is based on the direct environment.
human factors involved—with the motivation to influence the rules
& regulations devised around these human experiences. Additionally, Definition 4. 𝑣𝑖 ↦ 𝑛𝑖 |𝑣𝑖 ∈ 𝑉 ∧ 𝑛𝑖 ∈ R3 .
the parameter objects described in [6,78] could be extended using the
graph information generated within this paper and categorize a given In the following sections, a reference to 𝑛𝑖 is to reference the 3-
path as listed in Table 1. dimensional point of 𝑣𝑖 ∈ 𝑉 (𝐷), and similarly, a line segment 𝜇𝑖𝑗 ↦
𝑒𝑖𝑗 |𝜇𝑖𝑗 = 𝑛𝑖 𝑛𝑗 .
4. Accessibility graph method A set of geometry objects 𝛤 represents the interactable surfaces
relating to human accessibility.
The resulting graph is a mapping of graph vertices to point locations
in the environment, referred to as nodes. First, an interpretation of the Definition 5. 𝛤 ∶= {𝐵𝑖 … 𝐵𝑖+1 }.
3D model must be done in order to find the nodes accessible from a
starting point (Fig. 7). Once this mapping has been created, for each The set 𝐵𝑖 represents a collection of geometry objects with a par-
node in the environment, any human based analysis can be performed ticular property. For any generated graph 𝐷, |𝛤 | > 0 for at least
and attributed to that location. Considering the nodes in isolation, a one traversable surface must be defined (all objects can default to
visibility graph can be generated [52]. More complex and individual traversable). Additional surfaces can include obstacle objects as well
factors can also be analyzed, such as falling injury probability [79]. as specific surface types. For example, using BIM: hardwood or carpet
As with any building-based graph, the shortest path between two flooring could be defined, or with GIS: sidewalk concrete and grass
nodes can be found. When generating a graph based on a specific could be included.
6
Fig. 9. The parameters for traversing the model during the graph creation. Cross-slope
𝑠𝑐 is in the perpendicular plane to the figure, as illustrated in Fig. 3).
Fig. 8. The relationship of vertices and edges in the (𝑥𝑦) plane. Each color group → (𝑣 ) ∶= {𝑒 | 𝑣 ∈ 𝑁 + (𝑣 )}.
Definition 7. 𝐸𝐷 𝑖 𝑖𝑗 𝑗 𝐷 𝑖
(gray, blue, and red) represents the edge induced subgraph 𝐷[𝐸𝐷→ (𝑣𝑖 )]. The parent–
child relationship is defined as the from and to nodes, e.g., 𝑣1 is the parent and 𝑣5 is This set of outgoing edges is then used to define the edge-induced
a child of 𝑣1 . (For interpretation of the references to color in this figure legend, the subgraph 𝐷[𝐸𝐷 → (𝑣 )] of 𝑣 , which is seen in Fig. 8 through different
𝑖 𝑖
reader is referred to the web version of this article.)
colorings.
As ∀𝑣𝑖 ∈ 𝑃 , ∃ 𝐷[𝐸𝐷 → (𝑣 )], the weights associated with an edge
𝑖
𝑒𝑖𝑗 ∈ 𝐷[𝐸𝐷 → (𝑣 )] can be influenced by both node and edge values of
𝑖
4.1.2. Subgraph and subsets of locations its subgraph so that human-relevant factors such as cross slope can be
When discussing the graph in terms of accessibility, the basic con- accounted for.
cept is given as: from a given node, what is the cost, if possible, to get
to the next node? The graph is built such that the only nodes to move 4.1.3. Graph creation parameters
to, from any particular node, are ones that were checked specifically The graph generation results in the defined locations of accessibility
in relation to that node, which requires a unique check for each time a given specific parameters. These parameters are shared among many of
node is defined to have an accessible path to another node (Fig. 8 uses the navigation mesh algorithms (e.g., [46]) and while defaults can be
colored edges to illustrate connections between nodes). In other words, used, are input by a user. Parameters used within this algorithm are:
a parent node is the starting (initial) node (𝑛1 ) of a directed edge of
an ordered pair of nodes (𝑛1 , 𝑛2 ) and the child is the ending (terminal) 𝜏 : Starting Point (x,y,z)
node (𝑛2 ). This ordered pair (parent,child) is later abbreviated as (𝑝, 𝑐). ℎ : Height
However, discussed in the following sections, the use of neighboring 𝑎 : Grid Spacing (x,y)
nodes defines edge costs for cross-slope costs. 𝑏𝑢 , 𝑏𝑑 : Step Height (Ascent +, Descent -)
Without additional processing of the graph, it may be weakly con- 𝑠𝑢 , 𝑠𝑑 , 𝑠𝑐 : Slope (Positive +, Negative -, Cross)
nected, as a graph may have a directed edge to a node that has no 𝛾 : Minimum Required Children
outgoing edges. This may happen if the SHAPE is generated with areas 𝛷 : Set of Neighboring Directions
that allow an occupant to get to a location but not back (e.g., a ramp Fig. 9 illustrates the decision parameters from a given node (shown
that is feasible to traverse down, but not up). A separate issue is in green) to another. The slope and step parameters are calculated
when a node is defined as a child through the given parameters and separately. Slope is defined as an increase or decrease of height from
subsequently has no children. For this case, a specific parent set 𝑃 ⊆ 𝑉 one node to another when no geometry obstructs a direct path between
of nodes is defined. nodes. A step is when an obstruction of the path between nodes exist.
Uniquely, 𝛾 is the minimum children required to consider a node valid,
Definition 6. 𝑃 ∶= {𝑣𝑖 |(𝑣𝑖 , 𝑣𝑗 ) ∈ 𝐸}. and 𝛷 are the direction vectors to check for valid children such that
a set of 8 directions (0 and ±1) would be the adjacent 8 nodes while
As each node in 𝑃 has an outgoing edge, there is at least one
+ another direction set (e.g., 0, ±1, ±2) may include a secondary level
child node associated with it. The neighborhood 𝑁𝐷 (𝑣𝑖 ) is the set of
of subsequent nodes. Given 8 neighboring nodes, a 𝛾 value of 8 would
vertices dominated by vertex 𝑣𝑖 in graph 𝐷. While dense graphs are
mean a node is only valid in the graph if it is also able to access all
sometimes scrutinized for the space complexity, this is most often seen
directions. This check would fail if a node is close to a wall, creating a
in fully connected visibility graphs and not applicable here as edge
minimum buffer region.
connections are limited to a predefined number of immediate neighbors
These components allow for a great variety of human related acces-
(in euclidean space) of 𝑣, creating a sparse graph. Fig. 8 shows the base
sibility issues to be incorporated in the graph generation. For example,
case in which only adjacent points on the grid are possible child nodes
slope can be defined as an incline of 12:1 while the minimal step height
(an extension of this is described in the following sections).
+
can be 0 cm. In this case, there is a rule describing an allowable
For clarity, 𝑁𝐷 (𝑣𝑖 ) is shortened and referred to as the child set 𝐶
height difference when traveling between two locations; as long as
when the context is of the general node 𝑣. Next, a set of outgoing edges
there is a direct connection between those locations (i.e., it is a ramp
from a vertex 𝑣𝑖 is defined as: or incline, not an obstacle or step), describing the graph required of
a wheelchair. Parameters can be set at some maximal human capable
2
This can be seen in Appendix A and the supplemental material when a level (maximal clearance) with the values stored within multiple graphs
ramp is too steep or steps are not usable. as edge costs and later iterated through for various path-finding and
7
Algorithm 1: Build digraph of nodes and edges

1 𝑄←𝜏
2 𝑉 ←𝜏
3 𝐸, 𝑊 ← ∅, ∅
4 𝐷 ← {𝑉 , 𝐸, 𝑊 }
5 Function buildGraph(𝐷, 𝑄):
6 if 𝑄 = ∅ then
7 return 𝐷
Fig. 10. Overview of the algorithm to generate the accessibility-based graph. 8 𝑝 ← 𝑄.𝑝𝑜𝑝()
Rectangles (blue boxes) are the methods detailed in this section. 9 𝐶, 𝑤𝑡 ←getNodes(p)
10 𝑊 ← 𝑤𝑡
11 for c ∈ 𝐶 do
accessibility checks. Alternatively, a single graph can be built with 12 𝑊 ← setEdgeCost(𝑝, 𝑐)
the edge costs incorporated on-the-fly. As a note, post-calculating edge 13 𝑉 ←𝑐
costs is embarrassingly parallel as this is done ∀ 𝑣𝑖 ∈ 𝑃 , where 14 𝐸 ← (𝑝, 𝑐)
∃ 𝐷[𝐸𝐷→ (𝑣 )].
𝑖 15 if 𝑐 ∉ 𝑃 then
16 𝑄.𝑖𝑛𝑠𝑒𝑟𝑡(𝑐)
4.2. Algorithm overview 17 return buildGraph(D,Q)
In Fig. 10, a simple overview of the algorithm to build the

accessibility-based graph is shown with four main functions described
in the following sections. The algorithm initializes with the user pa- Two main functions are defined that determine the validity of a
rameters and geometry objects that remain in a global scope. Overall, child node within the getNodes() function (called in Alg. 1 and later
the algorithm iterates through a queue of nodes that are continually defined in Alg. 2). The function inter(𝑛𝑖 , 𝑛𝑗 ) is defined as returning the
added to the queue in a breadth-first manner from the starting location distance 𝑛⃖⃖⃖⃖⃖⃖ 𝑗 intersects 𝐵𝑖 in the environment and dst(𝑛𝑖 , 𝑛𝑗 ) returns
𝑖 𝑛⃗
𝜏 within the environment (buildGraph). While checking each item ‖𝑛⃗𝑖 − 𝑛⃗𝑗 ‖.
of the queue, getNodes iteratively finds the valid neighbor nodes by
Fig. 11 shows the node evaluation in the (𝑥𝑧) plane where the
calling checkChild on each potential neighbor. The final function
ray cast direction is −𝑧. ̂ As the graph is in relation to physical space,
in the process, getConnection, determines if an edge can exist
−𝑧̂ corresponds to the direction of gravity. The parameters and labels
between two nodes based on the input parameters, and if so, the type
used are those described in 4.1.3. The surfaces illustrated provide
of movement required between the original and potential child node.
examples of a step, faceted surface, and natural topography (e.g., a hill
Once an edge is determined to be possible, the movement type is
returned (likewise for all connections for the original queued item). or mountain). When a possible node is invalid, it is not added to 𝑄,
Before the breadth-first process continues to the next item in the queue, and in the simple example within the figure, the graph completes. To
the graph is updated with vertices, edges, and weights. simplify the illustration, Fig. 11 is a two-dimensional example is given
and only the directed relation from the left to right is discussed.
4.2.1. Breadth-first process In 11(a), a ray is cast in −𝑧̂ starting at 𝑛1 . As the node found from
The generation of common surface and graph information methods inter(𝑛3 , 𝑛3 − 𝑧) ̂ < 𝑏𝑢 , 𝑛3 is considered valid. When checking 𝑛4 , ℎ is
regarding navigation takes into account the entire environment at reset based on the previous node (𝑛3 ). To demonstrate the use of 𝛤 ,
once. Breadth-first search allows for the fast completion of the graph three objects, 𝛤 = {𝐵1 , 𝐵2 , 𝐵3 } are represented in the illustration where
generation when no accessible locations are found (Appendix A). This they are all walkable surfaces. The last node, 𝑛7 , is shown in red as
feature is particularly useful in the planning and design phase of a inter(𝑛7 , 𝑛7 − 𝑧) ̂ > 𝑏𝑢 , ending the graph generation. In the down step
building or environment as it offers immediate numerical and visual case for parameter 𝑏𝑑 , 11(b) illustrates the intersection point for 𝑛5
feedback about the features that are blocking accessibility. The quick falling outside the threshold. Finally a node 𝑛7 is invalid in 11(c) as
ending to the graph generation benefits the calculation time in practical the angle created from 𝑛6 and inter(𝑛7 , 𝑛7 − 𝑧) ̂ > 𝑠𝑢 . Furthermore, the
use cases (e.g., the starting location being a ramp that is too steep). relative positioning in 𝑍 of parent–child nodes can be seen by the
Generating the graph begins with a valid starting location where height variation relative to the previous node.
start point 𝜏 is above a valid surface geometry in 𝛤 . The algorithm Algorithm 2 implements this process in three-dimensions. The func-
begins with an empty graph 𝐷 and an ordered set 𝑄 representing tion getNodes(p) takes an input node 𝑝 and tests for its children in the
the queue of vertices that remain to be checked for validity in 𝑃 . At directions contained in a given set of an even grid in the (𝑥, 𝑦) plane
initialization 𝑄 = {𝜏}, with the algorithm completing when 𝑄 = ∅.
(i.e., Fig. 8), represented in the set 𝛷 = {−1, 0, 1} × {−1, 0, 1}. With
Generating the graph can be seen in Algorithm 1 line 8, where
these directions, a 45 deg angle is a limiting factor for a path connecting
a parent node 𝑝 is removed from the queue 𝑄. The valid children
nodes. To decrease this angle, 𝛷 can incorporate additional locations,
nodes are assigned to set 𝐶 and if a child node is not already defined
i.e., (−2,−1), which would correspond to node 𝑣3 connecting to 𝑣8 in
as a parent node, it is added to 𝑄. Once the set of valid children
Fig. 8. An alternative approach to expanding the neighboring nodes
are returned, the weight vector containing step type is stored in the
could be to iteratively expand while some condition remains true, such
weight set (line 10). This pseudocode shows the method in which
weights are calculated during graph construction, and therefore, during as no stepping, through an adaptive highway graph [56]. Algorithm 1
the iteration over the child set, a function setEdgeCost() on line 12 and 2 describe the three-dimensional breadth-first process in general
calculates a directional edge cost between the two nodes and stores it in terms. On line 4, a point 𝑐 is composed of the 𝑥, 𝑦, 𝑧 values of 𝑝, offset
𝑊 . This function can be used to set multiple types of edge costs, such as in the directions of 𝛷.
distance, slope, and energy, as discussed later. After the directed edge The process for determining the return value of checkChild(p,c)
of a parent and child is stored, the child node is checked for existence begins with the description of Fig. 11 and is expanded to define edge
in the parent set. If the node is not in the existing parent set it is added weights and determine criteria for an edge connection in the following
to 𝑄. section.
8
Fig. 11. Illustration of three common occurrences in graph creation of non-flat surfaces for checking a valid child node. Green circles represent a valid node and red circle as
invalid based on the parameters given. The black circle represents the starting node 𝑛1 and each gray circle is a subsequent check for a valid node. 11(a) represents climbing stairs
and multiple geometry objects, 11(b) is descending stairs, and 11(c) is a natural topography. The letters 𝑎, ℎ, 𝑏 represent the offsets used as values in the graph creation. 11(a)
shows the node numbering scheme with the same ordering followed in 11(b) and 11(c). (For interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
Algorithm 2: Check for valid neighbors Algorithm 3: Check if a child is within constraints
1 Function getNodes(𝑝): 1 Function checkChild(𝑝, 𝑐):
// Check parent 𝑝 for valid children // Check if child 𝑐 is a valid child
2 𝐶, 𝑤𝑡 ← ∅, ∅ 2 𝑐 ′ ←inter(𝑐, ⟨𝑐𝑥 , 𝑐𝑦 , 𝑐𝑧 − 1⟩)
3 for (𝑖, 𝑗) ∈ 𝛷 do 3 if 𝑐 ′ ∉ 𝐵𝑤 then
4 c ← (𝑝𝑥 + (𝑖 × 𝑎), 𝑝𝑦 + (𝑗 × 𝑎), 𝑝𝑧 + ℎ) 4 return ∅, ∅
5 𝑡, 𝑐′ ←checkChild(p,c) 5 𝑡 ←getConnection(p,𝑐 ′ )
6 if 𝑐′ then 6 if 𝑡 ≠ 𝑡INVALID then
7 𝐶𝑖𝑗 ← 𝑐′ 7 if 𝑡 = 𝑡DIRECT then
8 𝑤𝑡𝑖𝑗 ← 𝑡 8 if 𝑠𝑑 < (𝑝𝑧 − 𝑐𝑧′ ) < 𝑠𝑢 then
9 if |C|<𝛾 then 9 return 𝑡, 𝑐 ′
10 return ∅, ∅ 10 if 𝑏𝑑 < (𝑝𝑧 − 𝑐𝑧′ ) < 𝑏𝑢 then
11 else 11 return 𝑡, 𝑐 ′
12 return 𝐶, 𝑤𝑡
queries of the relationship between nodes (such as defining edge cost)

4.2.2. Determining accessibility returning both a height offset and direct connection allow for a specific
As discussed in the previous section, each (𝑝, 𝑐) is checked against categorization such as an incline instead of a step. This method provides
+ a similar but more detailed approach to determining node connections
a set of parameters and criteria. 𝑁𝐷 (𝑝) is constructed in Alg. 2, repre-
sented by the returned set 𝐶. One aspect of accessibility is a minimum as the edge culling in [54] or the edge connectivity determination of a
width or region a valid path can traverse through. Algorithm 2 can visibility graph [52,82].
be expanded to force accessible edges in path-finding to only pass Given the first connection type being false, a second ray, referred
through nodes that are connected to at least some user-defined number to in the figure as 𝑑2 , is calculated and used to check for a connection.
of neighbors. On line 12 of Alg. 2, rather than returning 𝐶, a check can Depending on the height relationship between 𝑝 and 𝑐, the translated
be done to see if |𝐶| is greater than some minimum connectivity and if start position of the ray is determined by 𝑏𝑢 or 𝑏𝑑 as the directed edge is
not, return an empty set. For example, if 𝛷 is the set of 8 neighboring either stepping up or down. The special case is for when the two nodes
directions, return 𝐶 ⟺ |𝐶| = 8. are equal in height with an obstruction between them, requiring a step
For 𝑐 ∈ 𝐶, 𝑐 is determined to be valid by function checkChild(p,c), over the obstruction (Fig. 12(b)), which has a ray start point calculated
shown in Alg. 3. First, the distance to the closest 𝐵𝑖 ∈ 𝛤 is found by the same as a step up.
inter(𝑐, 𝑐 − 𝑧).
̂ Given no labeled geometry in 𝛤 , 𝐵𝑤 is the same set Algorithm 4 demonstrates the process for calculating the rays and
of objects as 𝛤 . For labeled geometry consisting of walkable and non- connection types. After the connection type and parameters of 𝐶 is
walkable surfaces, a set of walkable surfaces are defined as 𝐵𝑤 ⊂ R3 . If determined, the algorithm returns to buildGraph(D,Q) on line 5. At
the closest distance is to 𝐵𝑖 ∉ 𝐵𝑤 the node is not valid. The connection that point, for the given 𝑝 passed to getNodes, the subgraph of 𝑝 can
+
type 𝑡 between 𝑝 and 𝑐 ′ is calculated by getConnection(p, 𝑐 ′ ) and if be constructed (𝐷[𝐸𝐷 (𝑝)]).
valid, the step and slope parameters are verified. In implementation,
the relational information of connection types between the parent and 4.3. Edge costs
child nodes is stored during the graph construction, and the function
returns the connection type (𝑡) and valid node (𝑐 ′ ). The cost of traveling from one node to another is defined by the
The type of connection between two nodes is determined as one mapping of the weights from set 𝑊 as shown in Section 4.1.1. While the
of five possible cases (Fig. 12): No Step 12(a), Step Over 12(b), Step significance of the method and resulting graph is the straightforward
Up 12(c), Step Down 12(d), Invalid 12(e). The reference to ‘‘No Step’’ application to a variety of human factors, a few weight categories pre-
is specifically regarding a geometry that protrudes off of the ground viously discussed in the literature will be defined here. Especially at the
surface, which would require a foot, wheelchair, or other aid to lose urban scale, the inclusion of varying ground conditions to walkability
contact with the ground during traversal. The ‘‘No Step’’ is referred to analysis can better predict the human experience, with comfort, timing,
as a direct connection, which works in both flat and inclined ground and fatigue being realized.
as the height offset between nodes is also stored. Therefore, future Two cases exist for defining the cost of an edge:
9
Fig. 12. Example configurations for a connection between nodes. First, a check 𝑑1 is performed (12(a)). When the first check is not successful, depending on the relation between
the nodes, either the up (12(c)) or down step (12(d)) limits are used for a second check 𝑑2 . If 𝑐𝑧 = 𝑝𝑧 , a special case is defined (12(b)). In (12(e)), the incline after the parent and
down step of the child block a connection. This last case is the same result given a wall or another obstacle preventing an edge between the nodes.
Algorithm 4: Check for node connection type The requirement for 𝛹𝑡 (𝑒𝑖𝑘 ) = 𝑡𝙳𝙸𝚁𝙴𝙲𝚃 corresponds to the connection
type defined in Alg. 4 line 3 as a direct connection with no step.
1 Function getConnection(𝑝, 𝑐):
// Check parent 𝑝 for connection type to 𝑐 Additionally, a function defining energy expenditure3 of walking on
2 if dst(𝑝,𝑐) ≤ inter(𝑝,𝑐) then slopes [13] as shown in Fig. 4 is used for 𝑤𝑒 . By correctly weighing
3 return 𝑡DIRECT the energy of moving both up and down a slope, human based smooth
4 if 𝑝𝑧 ≤ 𝑐𝑧 then paths, rather than mathematically defined ones (e.g., [64,65]), can be
5 𝑝𝑧 += 𝑏𝑢 found.
6 if dst(𝑝,𝑐) ≤ inter(𝑝,𝑐) then
7 if 𝑝𝑧 = 𝑐𝑧 then Definition 9. 𝛹𝑒 (𝑒𝑖𝑗 ) = 280.5𝛹𝑠 (𝑒𝑖𝑗 )5 − 58.7𝛹𝑠 (𝑒𝑖𝑗 )4 − 76.8𝛹𝑠 (𝑒𝑖𝑗 )3 +
8 return 𝑡OVER 51.9𝛹𝑠 (𝑒𝑖𝑗 )2 + 19.6𝛹𝑠 (𝑒𝑖𝑗 ) + 2.5.
9 return 𝑡UP
Edge costs are reasonable, but often distracting or difficult to visu-
10 if 𝑝𝑧 > 𝑐𝑧 then
alize in an environment. It is therefore beneficial to attribute values to
11 𝑝𝑧 += 𝑏𝑑
a node that represent the costs associated with it. Likewise, a node may
12 if dst(𝑝,𝑐) ≤ inter(𝑝,𝑐) then
be an input mechanism for different types of simulations. In this case,
13 return 𝑡DOWN
the node value can be a combination of all directed edge costs away
14 return 𝑡INVALID → (𝑣 ))), as a localized decision making agent
from the node (i.e., W(𝐸𝐷 𝑖
(e.g., chemtrails used in [40]) can be directed away from nodes that
would have been difficult to leave from. The value of this node can be
1. Cost of mobility along the edge represented as:
2. Cost of moving to a node ∑
𝜄 ⊙ 𝜌𝜄
→ (𝑣 ))
𝜄 ∈𝑊 (𝐸𝐷
In the first group, metrics related to mobility itself are included; 𝑖
𝛹𝑐 (𝑒𝑖𝑗 ) = Cross-slope, 𝛹𝑑 (𝑒𝑖𝑗 ) = Metric Distance, 𝛹𝑠 (𝑒𝑖𝑗 ) = Slope, 𝛹𝑒 (𝑒𝑖𝑗 )

Where each weight has corresponding (user defined) coefficient
= Energy Expenditure, 𝛹𝑡 (𝑒𝑖𝑗 ) = Step Type.
vector 𝜌 of importance.
In the second group, attributes not able to be calculated by the
Another case of defining an edge cost is when the existing attributes
path in isolation, but rather as values calculated cohesively with the
built environment are considered. These highly localized parameters or values are first associated with a node, rather than an edge, defined
are easily missed in maximal edge connected graphs (e.g., [58]) and for when an edge 𝑒𝑖𝑗 is assigned a weight equal to either its initial
rely on the dense set of nodes throughout the environment. In this vertex (𝑣𝑖 ) or terminal vertex (𝑣𝑗 ). This often occurs when a value from
approach, an edge weight is defined by some properties or attributes of simulation or evaluation, such as lighting, is mapped to locations in the
the end points (nodes) of the edge. For example, a directed edge with a environment, rather than the cost of moving between locations.
start node located on concrete and end node located on sand is given a
higher cost than the reverse as that node should be avoided. This weight 4.3.1. Path costs
mapping can include 𝛹𝑘 (𝑒𝑖𝑗 ) = Type of Surface the end node is on, with
The focus of this paper remains in the graph generation. In this
further types; 𝛹𝑙 (𝑒𝑖𝑗 ) = Light Intensity from Opposite Direction, 𝛹ℎ (𝑒𝑖𝑗 )
section, defining various combinations and evaluations of a path is
= Light Difference Between Nodes, 𝛹𝑓 (𝑒𝑖𝑗 ) = Probability of Falling,
𝛹𝑣 (𝑒𝑖𝑗 ) = Percentage of View to Greenspace, 𝛹𝑤 (𝑒𝑖𝑗 ) = Surrounding briefly discussed but is not exhaustive due to the numerous possible
Width. combinations. While a path is often defined as a sequence of vertices, a
Some edge cost definitions such as the metric distance, 𝛹𝑑 (𝑒𝑖𝑗 ) = modified definition is used here to provide a simple connection to the
|𝑛𝑖 − 𝑛𝑗 |, are straight forward while others 𝛹𝑙 (𝑒𝑖𝑗 ) are based on the weights described in Section 4.3. As a note, the vertices of the graph are
implemented simulation technique. The two most notable costs related associated with nodes in the environment and therefore, the methods of
to the focus of this paper are 𝛹𝑐 (𝑒𝑖𝑗 ) and 𝛹𝑒 (𝑒𝑖𝑗 ) and are therefore path evaluation seen in space syntax such as angle or number of turns
defined in full. 𝛹𝑐 (𝑒𝑖𝑗 ) is the mapping of weights to edge costs in the is straightforward to use. However, this section focuses on path analysis
graph relating to the cross slope and 𝛹𝑡 (𝑒𝑖𝑗 ) is the mapping consisting based on edge weights.
of an identifier for the step type found in Alg. 4. A directed path 𝛺 is an ordered sequence of edges that are con-
Calculating the cross-slope score 𝛹𝑐 (𝑒𝑖𝑗 ) is done by considering the nected in 𝐷 between two vertices.
edges to 𝑢𝑖𝑗 (using the mapping defined in Section 4.1.1) that are
orthogonal in the 𝑥𝑦 plane, denoted by 𝑥𝑦 ⊥. The maximum difference
Definition 10. 𝛺 ∶= (𝑒1 , … , 𝑒𝑚 ∣ 𝑒𝑛 ∈ 𝐸(𝐷)).
in height between the nodes of orthogonal edges is used as the score:
→ (𝑣 ) ∧
Definition 8. 𝛹𝑐 (𝑒𝑖𝑗 ) = 𝚖𝚊𝚡(|𝑧̂ ⊙ (𝑛𝑗 − 𝑛𝑘 )| ∀ 𝑢𝑖𝑗 𝑥𝑦 ⊥𝑢𝑖𝑘 ) ∣ 𝑒𝑖𝑘 ∈ 𝐸𝐷 3
𝑖 This is one example, although other methods can be easily included or
𝛹𝑡 (𝑒𝑖𝑘 ) = 𝑡𝙳𝙸𝚁𝙴𝙲𝚃 . used in place.
10
Fig. 13. In 13(a) the nodes defined as accessible locations are visualized in green. Three kitchen layouts (13(b), 13(c), 13(d)) are all analyzed with identical graph and path
algorithms and no modification to the parameters with a path between the sink, stove, and refrigerator. 13(b) shows a typical work triangle of a kitchen. In 13(c), the path
between the refrigerator and sink is interrupted by an island, diverting the path around it. 13(d) shows a smaller and modified layout where the path between the key points is
nearly a straight line. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Defining the cost of an edge directly impacts the paths found free-form design and planning stages where large changes to a de-
when applying any common shortest path algorithm, which is synony- sign based on overall accessibility can be easily made. While the
mous with least-cost in this context. If considering the combination same methods can be implemented in BIM focused software (e.g., Au-
of distance and energy in a weighted function, this path in an urban todesk Revit), the demonstration within a modeling package containing
environment can be used as the starting point for defining ramps or limited information on the geometry demonstrates the scaleable use.
walkways to be constructed as to minimize the amount of land mass The code was developed in Python and IronPython, with a SQL
that must be removed. The breadth-first approach provides an efficient database to manage graph data between them. During the graph cre-
method in calculating this aspect of spatial evaluation as the graph ation a Python Dictionary object is used to store the vertex and edge
generation ends when one of the accessible parameters is not met. In connections, which is later converted to a compressed sparse row (CSR)
the indoor environment, these paths of least resistance can give direct structure with multiple arrays set as data for costs. To demonstrate
feedback on the performance of a layout or distance between rooms. the algorithms in various cases a simple GUI was made with the
When considering the value in quantifying occupant paths of a space, Rhino Grasshopper interface. No modifications to the algorithm are
multiple approaches of cost aggregation can be useful, such as distances made between examples, except for any human factor input parameters
with types of spatial views. that are used for demonstration purposes. The combination of these
By combining the weighting parameters discussed in Section 4.3, examples illustrates the flexibility in design use-cases and robustness
the shortest path between two buildings or rooms can be quantified by at varying levels of geometric detail.
more than simply able-bodied walking distances. As design alternatives While more concrete examples of the algorithm are given in the
and modifications are made to the environment, these values update following sections, Fig. 2 provides an overview of the process from a
accordingly and enable a quantified score of evaluation across the design-user perspective. Given an initial building model, the user would
designs. When separating the costs, multiple paths can be generated select a location to start the algorithm from. Selecting the metrics
for each one (or any combination there of). A space could therefore be of interest, such as the open space around each location, a heatmap
evaluated based on the variance between each of the paths. visualizing the data can be applied to the nodes of the graph. Either
Additionally, edge costs can be used as the evaluation metric where, the nodes themselves can be explored, or the nodes with the building
for example, the metric distance is used as the edge cost to generate model can be seen in the modeling program at the same time (or nodes
least-cost paths between desired locations. Next, the weight of each can be hidden all-together). Each image in the diagram is taken as a
screen capture from the viewport, with only the text box and start/end
edge, or the combination of the weight vector values, is applied as the
waypoints overlaid for illustration purposes.
evaluative score of a path, such that 𝑆 ∶ 𝛺 → R.
To demonstrate the feasibility, a SHAPE was generated on three test
cases: a kitchen representing the object scale, the NJIT Architecture
Definition 11.
∑ building for the building scale, and two height varying surfaces rep-
𝑆(𝛺) ∶= 𝑤𝜄 ⊙ 𝜌. resenting terrain types. The intent of these examples is to clearly and
𝑤 𝜄 ∈𝑊 (𝛺) understandably communicate the flexibility, performance capability of
In some cases the weight vector must be first processed, or alterna- the algorithm presented in this paper, and the direct impact quantita-
tively calculated, to have a meaningful result. In the case of a step, 𝑤 tive metrics have on the generated graph. Both the kitchen (Section 5.1)
contains an enumerator associated with connection types (Alg. 3 line and building (Section 5.2) models are exported from an Architects
5). Therefore to count the total number of steps required in a path, the Revit and Interior Designers Rhino 3D models (Adapted from [79]).
index of connection type 𝑡 in the weight vector is redefined as: In Section 5.3, both a real-world GIS terrain is shown and a more
{ simplified model is used to help readers in understanding how varying
1 if 𝑤 [𝑡] = 𝑡𝙾𝚅𝙴𝚁 ∨ 𝑡𝚄𝙿 ∨ 𝑡𝙳𝙾𝚆𝙽 terrain types impact energy costs and can be included in path planning
𝑤 [𝑡] =
0 otherwise and spatial analysis.
To count the number of steps in a path, 𝜌 would be a vectors of 0’s, 5.1. Demonstration on room scale
with the exception of the connection type coefficient being 1.
At the room scale, a kitchen provides a demonstration of the graph
5. Implementation avoidance of objects. A common design process for a kitchen is the
evaluation of a work triangle defined by the refrigerator, sink, and
The algorithms described were implemented in Python and vi- stove (store, preparation, and cooking). By selecting these objects as
sualized in the Rhino 3D software as it is commonly used in the destinations for a path, this triangle can be generated and evaluated
11
Fig. 14. Demonstration of different paths created with the same graph using different and combined attributes. 14(a) shows a heatmap visualization of the average distances in
a viewshed analysis, where red is the most open views and blue the least open views. 14(b) is a standard shortest path using metric distance. 14(d) avoids paths based on the
maximum view while 14(c) is the inverse. In comparison, 14(e) avoids edges based on the minimum view distance. Finally, 14(f) combines the edge costs of open space (14(e))
and distance (14(b)) by setting a multiplier to edges within a threshold close to obstacles, creating a wall-buffer effect. (For interpretation of the references to color in this figure
legend, the reader is referred to the web version of this article.)
automatically. The object level inclusion of the graph enables kitchen solve only the pathing similarity problem. The importance of using data
islands to be included in the evaluation rather than considering the to determine a path or evaluating a shortest path was elaborated on in
direct path between the objects of interest (e.g., stove). Section 2. By predetermining how occupants will move or experience
Using the accessibility graph, Fig. 13 shows three basic kitchen an environment, the evaluation of that movement loses meaning. While
layouts. The table and chairs, along with counters are not included these options are the best (only) ones available, the ongoing efforts
in the accessible locations. The stove, refrigerator, and sink positions of human-building interaction and cognitive science enables us to
are selected to calculate the travel between them, creating the work more accurately model and simulate environmental conditions of an
triangle. Using the distance metric of the path, the variation in the occupant.
triangle perimeters is simply calculated. As the path is created from the By no means is spatial visibility and distance the only two conditions
nodes, an exact shortest path can vary based on the distance a node is considered by an occupant. However, Fig. 14 demonstrates the diversity
connected to its neighbors. In Fig. 13(b) the total calculated distance is of paths between identical locations when only two quantitative metrics
1091 cm, while the exact distance is 1080 cm. The difference between are used in the SHAPE. In this demonstration, a viewshed analysis is
the estimated and actual distances is 11 cm, or 1%. Individual paths are performed at a height of 1.8 m using 2000 rays per location and a 360
calculated as follows (calculated vs actual): Stove–Refrigerator 242 cm degree field of view, with ±40◦ elevations. The SHAPE was generated
vs 241 cm, Stove–Sink 343 cm vs 340 cm, Sink–Refrigerator 506 cm vs at a grid spacing 𝑎 of 25 cm with a minimum connectivity of 8 nodes,
500 cm. which the result can be seen in the top right empty rooms (Fig. 14(a))
In comparison, to Fig. 13(b), an island blocking the direct that have an opening too small to be considered accessible. For each
(Fig. 13(c)) path changes the total distance from 1091 cm to 1114 cm. location, the maximum and minimal distance any ray travels in the set
With the island blocking the path, the euclidean distance between is recorded and used as node attributes. These values are then set as
the sink and refrigerator is calculated as 532 cm compared to the the outgoing edge cost (incoming costs would provide similar results)
original 506 cm. Finally, the smaller kitchen arrangement of Fig. 13(d) to that node. The use of maximal or minimal views are not suggested
provides a total perimeter of 830 cm for the work triangle. While these as realistic paths, but rather as a demonstration of how these metrics
values represent an isolated test case for demonstrating the method in can find a variety of pathing strategies. Additionally, any metric can be
implementation, the example is in the context of a commonly defined calculated along any path.
approach to interior design. In the simplest form, this evaluation pro- In Fig. 14(b) the shortest path by distance between two locations
vides instant feedback as to the distances between objects in the work is visualized, totaling 78.7 m. Another example is to use the farthest
triangle in which designers may want to minimize for a desired occupant distance visible from a location as the edge cost, which would create
experience. In addition to traveling distances, a heatmaps could be high costs for any edge that has a long line of sight. This approach
visualized for the designer to highlight areas that are dangerous for creates the path in Fig. 14(d), which is 90.2 m long. The reciprocal
an occupant to fall, or locations difficult to reach certain cabinets, as 1
score ( 𝚖𝚊𝚡_𝚍𝚒𝚜𝚝 ), where edges cost less when there is a distant object
illustrated in [79]. (i.e., longer line of sight) is shown in Fig. 14(c) at length of 95.5 m.
Rather than avoiding or using edges based on the longest distance,
5.2. Demonstration on building scale the same can be applied for the shortest (i.e., closest distance to an
object in view). Similarly to the previous method, close distances can
When discretizing the model to nodes of accessible locations, certain be avoided by dividing 1 by the minimal distance, shown in Fig. 14(e),
features such as proximity to a wall are not accounted for when with a path length of 90.9 m. A final path is shown in Fig. 14(f) that
using a shortest path algorithm. In fact, this is one motivation for the combines the edge costs of distance and a function of minimal views(𝛺
importance of numerous quantitative human factors being included in in Section 4.3.1). By setting a threshold of 63 cm, any outgoing edge
the graph as the majority of existing work either: removes graph nodes from the node is set to the length of the edge multiplied by a constant
near obstacles to force paths away (i.e., the walls), uses an extremely (4, in this example). This is, practically, making an edge that moves
sparse graph containing predefined locations (e.g., center of doorways), within 63 cm of a wall cost four times as much as traversing an edge
or ignores the issue by interpreting the data as an approximation. When further away. By doing this, a minimal distance path is created that
removing nodes, the graph loses data that exists for human occupants to avoids obstacles (e.g., walls, handrails) without needing to remove
12
Fig. 15. (Top) Demonstration of using GIS data as input model for generating graph and analysis. Nodes are visualized as a heatmap from low to high energy cost (blue to red).
(Bottom) Paths are demonstrated from start-point (blue pin) to end-point (red pin) based on energy and distance. (For interpretation of the references to color in this figure legend,
the reader is referred to the web version of this article.)
Fig. 17. Two paths generated with (17(a)) distance only and (17(b)) energy.
Fig. 16. Two paths generated with (16(a)) distance only and (16(b)) distance with the Cross-slope avoidance in path generation is demonstrated in Fig. 16.
added cost of a cross-slope.
In Fig. 16(a), two corners of a ramp are used as start and end points
of the path, with a shortest path metric of distance being used. When
the angle of the cross-slope is added as an additional weight to the
nodes from the graph of accessible locations. This combined-cost path distance, the shortest path generated follows the ramps incline and
totals 82.2 m, 3.5 m longer than the minimal distance and 8.7 m shorter moves diagonally on the flat surface (Fig. 16(b)).
than the path purely avoiding obstacles. As the edge costs of various factors can be either the method for
finding the shortest path, or for generating it, there are two evaluative
5.3. Demonstration on topography scores. In 17(a), the shortest path is created using the edge costs of
distance in which the value is 64 m and the energy cost is 470 J/KG
At the largest scale, Fig. 15 shows a process in which GIS data was m. In 17(b), energy is used as the edge cost, reducing the energy
used to visualize areas of high and low energy expenditure, as well calculation to 355 J/KG m, with a distance of 73 m. This exemplifies the
complex relationship in landscape environments for human mobility as
as paths generated between locations that minimize either distance or
a 14% increase in path distance results in a 24% decrease in energy.
energy. In this example, a region near Park City, Utah was selected
Both the visualization and data can be used in the planning of
within CityEngine [83] at a size of approximately 2500 m2 . In this
environment paths and site selection. By understanding the physical
example, the limitation of the algorithm is not the size, but rather
impact the topography has on people, designers can include in their
the model resolution. The accuracy at which the topography can be
planning a quantitative metric of consideration for placing entrances
created limits the meaning of certain conditions, such as slope and step.
and exits in a way that may improve accessibility and reduce the need
For example, while there are clearly trees in the mountains, the 3D
for large changes to the natural environment.
model itself lacks these geometries. The geometry can still be analyzed,
but two conditions are applied: (1) The margin of error for specific 6. Discussion
metrics such as energy expenditure is assumed to be large (2) Spacing
of the nodes is sufficiently large so that errors from tessellated ground The consideration of fatigue and energy in an environment is greatly
conditions are minimized. In this example, node spacing was set to 5 applicable in urban environments and decisions as to the reduction
m, across which a maximum slope of ±45◦ was allowed. Looking at the of hills and soil, a costly construction process. The results show how
node visualization it becomes apparent there existing pathways that a highly complex topography can be analyzed and a minimal-energy
have been evaluated as low energy. Upon closer inspection, while the path (which has a non-linear relationship with slope), is automatically
node spacing is larger than a human step, roadways that have been found. Indoor, this type of calculation has important implications for
carved into the mountain are still captured at the course resolution large spaces such as airports, and with the integration of indoor and
and are found by the SHAPE. As the complexities of the large-scale outdoor, environments such as a school campus can use this type of
terrain are difficult to intuitively understand, two simple examples automated evaluation method for improving and determining accessi-
are additionally provided to clearly demonstrate the role of slope and bility and timing between classes (i.e., large scale built environment
cross-slope. circulation analysis).
13
The visualization method of heatmaps shown in Figs. 14(a) and metrics, this research, and corresponding in-depth literature review,
15 is a commonly used technique for providing intuitive feedback for will ideally reinforce the need for considering numerous human factors
a designer. Likewise, path visualizations (Figs. 13–17) are commonly in all aspects of computational design. Ideally, the increased use of
used in circulation analysis. In either technique for providing feedback machine learning and optimization methods for the built environment
to the designer, the important aspect is in what data is being visualized. and design will utilize evaluative data about people (i.e., variation in
By using the SHAPE method, not only can existing work (e.g., visibility floor surface and accessibility) and as such include the human more,
graphs) be used with its base graph structure, but high-fidelity human- not less, in the design of a space.
centric data relating to physiology can be at the forefront of the design.
With construction techniques pushing the limits of creation in the built 7. Conclusion
environment, such as creating curved concrete floors and ramps, it
is important to provide tools that can quantitatively evaluate how a
In this paper, a method for creating a weighted and directed graph
human will physically experience and navigate such a design. (digraph) from a 3D model with a focus on human-relevant attributes
There are multiple ways to approach the navigable or accessible for traversing a space, referred to as SHAPE, was presented. The
areas in a 3D model, each with their own drawbacks. While a pro- process for generating the graph uses ray-casting on the data structure
cess using voxelization could under-include geometry (e.g., ignoring used for other purposes in CAD, such as a Bounding Volume Hier-
geometry below the defined voxel size), it could likewise over-include archy (BVH), which improves performance and minimizes the time
these small details. As the method presented in this paper uses ray it takes for design-analysis. SHAPE was created within environments
intersections, a point of failure in accurate representation of accessible containing multiple levels and stairs, as well as on varying heights
locations could be a model containing extremely small geometry such such as slopes, cross-slopes, and other complex surface conditions,
as a metal wire across the room. Future work to resolve this could demonstrating unique paths based on distance, energy, and numerous
include the use of geometry bounding boxes relative to the edge visibility metrics—not feasibly found manually by a designer. Not only
connection of nodes. does SHAPE work on varying ground conditions, but includes the
The flexibility in the proposed algorithm for use on non-labeled data information as data attributes in the graph (e.g., node connections as
is similarly limited in instances where there is no guaranteed way to a step or incline) which enable high fidelity accessibility checks to
infer movable structures from the 3D model, such as the case of eleva- include physiological effort and other human-experiences.
tors. This is an inherit issue with all methods, and extending this work In summary, the following benefits to the integration of various
to utilize the same techniques (e.g., defining a node at the landing of design processes are achieved: (1) Does not require extensive pre-
the elevator on each floor) would be straightforward. The SHAPE could processing such as voxelization, making the query-result process faster.
be generated on each floor, with a labeled elevator (and hence, closest (2) Does not require (but can incorporate) labeled data such as BIM,
node) connecting the floors. Furthermore, the step height parameter enabling earlier-stage feedback (e.g., with mass models). (3) Creates a
could be set to 0, illustrating through path visualization an occupant Graph representation rather than general surface or polygon bounds,
using a wheelchair would have limited accessibility. It is still important enabling spatially high-fidelity metrics to be incorporated. (4) Only
to note, however, that the proposed method excels without labeled includes human-accessible locations in the graph, reducing computa-
tion time as well as highlighting areas of inaccessible locations. (5)
data and is able to traverse 3D environments on varying surfaces and
Works for both indoor and outdoor environments regardless of building
stepping blocks (e.g., stairs) even when these objects are not labeled
levels. (6) Defines and calculates edge costs by human factors, including
as such, which would likewise highlight to a designer that subsequent
slope, cross-slope, and steps — providing a higher resolution than past
floors were not accessible without elevators (i.e., missing a ramp).
works of occupant experience and enabling more accurate accessibility
Changing the user input parameters can allow for multiple specific
analysis for differently-abled occupants. (7) Easily used with existing
graphs to be analyzed, for example, useful for toddlers/children in
graph theory (e.g., visibility graphs) and design evaluation methods
childproofing a room by setting the height small enough for a child such as lighting as to not replace the vast existing literature, but instead
and see if they can access unsafe space, or detect sharp edges of objects enhance and augment it.
within their path, or give a metric for them climbing on furniture. Like-
This work can aid designers and researchers in automated analysis
wise, an extension to the weighting parameters of discrete evaluation of the built environment for physiological metrics of varying types of
metrics defined for edge costs, an additional dimension of the weight occupant mobility. SHAPE can be a platform for others to quickly and
vector could be added to incorporate time. This could be used, for easily generate graph representations of the space and integrate their
instance, in considering factors such as lighting levels where additional own additional metrics, useful in facilitating quantitative evaluations
edge costs would be generated across times of the day. of interest, and facilitating human-centric designs.
While this research focuses on presenting a method for generat-
ing a graph and then using the graph for path-based analysis, the
Declaration of competing interest
human factors discussed, and underlying graph representation, have
applications in related research areas. Although a navigation mesh
The authors declare that they have no known competing finan-
could be more efficient if the only goal was agent-based simulation,
cial interests or personal relationships that could have appeared to
the digraph could be used as a layer for both navigation policy and
influence the work reported in this paper.
as a position-aware data structure for storing information on agent
interactions (e.g., chemical diffusion rates for agent modeling [40]).
In furniture layout optimization procedures [84,85], an input metric Acknowledgments
is needed to drive the weighting factors of different configurations.
Although early work in this area used object relationships such as couch This paper contains work that was supported in part by the U.S.
to television [84], future research could leverage the graph representa- Army Combat Capabilities Development Command (CCDC) Armaments
tion embedded with human factor data at each node to include more Center and the U.S. Army ManTech Office under Contract Delivery
complex dynamics, such as visual noise calculated by the intersection Order W15QKN19F0002 - Advanced Development of Asset Protection
of multiple objects within a view frustum. Beyond discrete evaluation Technologies (ADAPT).
14
Appendix A
Please refer to https://doi.org/10.1016/j.autcon.2021.103557 for the full video of the graph generation.
Fig. A.1. Example stills of the supplemental video showing the growth of the accessible node locations. In A.1(c), the breadth-first search for accessibility has stopped at the steep
ramp on the right side due to the input user parameters, while subsequent frames of A.1(e) and A.1(f) show the graph continuing on the spiral ramp and connecting on the down
ramp. As the graph is directed a shortest path from bottom to top would use the spiral ramp, a path from top to bottom would use the ramp on the right.
References [16] B. Pierret, K. Desbrosses, J. Paysant, J.-P. Meyer, Cardio-respiratory and subjec-
tive strains sustained by paraplegic subjects, when travelling on a cross slope
[1] M. Frontczak, P. Wargocki, Literature survey on how different factors influence in a manual wheelchair (MWC), Applied Ergonomics 45 (4) (2014) 1056–1062,
human comfort in indoor environments, Building and Environment 46 (4) (2011) http://dx.doi.org/10.1016/j.apergo.2014.01.002.
922–937, http://dx.doi.org/10.1016/j.buildenv.2010.10.021. [17] A.C. Novak, S.M. Reid, P.A. Costigan, B. Brouwer, Stair negotiation alters
[2] F. Moura, P. Cambra, A.B. Gonçalves, Measuring walkability for distinct pedes- stability in older adults, Lower Extremity Review 2 (10) (2010) 47–51, https:
trian groups with a participatory assessment method: A case study in Lisbon, //lermagazine.com/article/stair-negotiation-alters-stability-in-older-adults.
Landscape and Urban Planning 157 (2017) 282–296, http://dx.doi.org/10.1016/ [18] A.C. Novak, V. Komisar, B.E. Maki, G.R. Fernie, Age-related differences in
j.landurbplan.2016.07.002. dynamic balance control during stair descent and effect of varying step geometry,
[3] A. Forsyth, What is a walkable place? The walkability debate in urban design, Applied Ergonomics 52 (2016) 275–284, http://dx.doi.org/10.1016/j.apergo.
Urban Design International 20 (4) (2015) 274–292, http://dx.doi.org/10.1057/ 2015.07.027.
[19] A.S. Voloshina, A.D. Kuo, M.A. Daley, D.P. Ferris, Biomechanics and energetics
udi.2015.22.
of walking on uneven terrain, Journal of Fish Biology (2013) 3963–3970, http:
[4] H.V. Krambeck, The Global Walkability Index (Ph.D. thesis), Massachusetts
//dx.doi.org/10.1242/jeb.081711.
Institute of Technology, 2006, pp. 1–135, http://hdl.handle.net/1721.1/34409.
[20] J.V. Jacobs, A review of stairway falls and stair negotiation: Lessons learned
[5] L.D. Frank, J.F. Sallis, B.E. Saelens, L. Leary, K. Cain, T.L. Conway, P.M. Hess,
and future needs to reduce injury, Gait & Posture 49 (2016) 159–167, http:
The development of a walkability index: application to the Neighborhood Quality
//dx.doi.org/10.1016/j.gaitpost.2016.06.030.
of Life Study, British Journal of Sports Medicine 44 (13) (2010) 924–933,
[21] D.S. Marigold, A.E. Patla, Age-related changes in gait for multi-surface terrain,
http://dx.doi.org/10.1136/bjsm.2009.058701.
Gait & Posture 27 (4) (2008) 689–696, http://dx.doi.org/10.1016/j.gaitpost.
[6] J. Shin, J.-K. Lee, Indoor walkability index: BIM-enabled approach to quantifying
2007.09.005.
building circulation, Automation in Construction 106 (2019) 1–14, http://dx.doi.
[22] B.W. Schulz, Minimum toe clearance adaptations to floor surface irregularity and
org/10.1016/j.autcon.2019.102845.
gait speed, Journal of Biomechanics 44 (7) (2011) 1277–1284, http://dx.doi.org/
[7] S. Gamache, F. Routhier, E. Morales, M.-H. Vandersmissen, N. Boucher, Mapping
10.1016/j.jbiomech.2011.02.010.
review of accessible pedestrian infrastructures for individuals with physical [23] P.C. Dixon, J.V. Jacobs, J.T. Dennerlein, J.M. Schiffman, Late-cueing of gait tasks
disabilities, Disability and Rehabilitation: Assistive Technology 14 (4) (2019) on an uneven brick surface impacts coordination and center of mass control in
410–422, http://dx.doi.org/10.1080/17483107.2018.1449018. older adults, Gait & Posture 65 (2018) 143–148, http://dx.doi.org/10.1016/j.
[8] S. Bennett, R.L. Kirby, S. Bennett, R.L. Kirby, B. Macdonald, Wheelchair ac- gaitpost.2018.07.168.
cessibility: Descriptive survey of curb ramps in an urban area, Disability and [24] A. Merryweather, B. Yoo, D. Bloswick, Gait characteristics associated with trip-
Rehabilitation: Assistive Technology 4 (1) (2009) 17–23, http://dx.doi.org/10. induced falls on level and sloped irregular surfaces, Minerals 1 (1) (2011)
1080/17483100802542603. 109–121, http://dx.doi.org/10.3390/min1010109.
[9] C. Eastman, J.-m. Lee, Y.-s. Jeong, J.-k. Lee, Automatic rule-based checking [25] A.M. Koontz, R. Cooper, M.L. Boninger, Y. Yang, B.G. Impink, L.H.V. van der
of building designs, Automation in Construction 18 (8) (2009) 1011–1033, Woude, A kinetic analysis of manual wheelchair propulsion during start-up
http://dx.doi.org/10.1016/j.autcon.2009.07.002. on select indoor and outdoor surfaces, Journal of Rehabilitation Research &
[10] US Access Board, ADA Accessibility guidelines (ADAAG), 2002, https: Development 42 (4) (2005) 447–458, http://dx.doi.org/10.1682/jrrd.2004.08.
//www.access-board.gov/guidelines-and-standards/buildings-and-sites/about- 0106.
the-ada-standards/background/adaag. [26] R.M. Glaser, M.N. Sawka, S.W. Wilde, B.K. Woodrow, A.G. Suryaprasad, Energy
[11] R.G. Soule, R.F. Goldman, Terrain coefficients for energy cost prediction, Journal cost and cardiopulmonary responses for wheelchair locomotion and walking on
of Applied Physiology 32 (5) (1972) 706–708, http://dx.doi.org/10.1152/jappl. tile and on carpet, Spinal Cord 19 (4) (1981) 220–226, http://dx.doi.org/10.
1972.32.5.706. 1038/sc.1981.45.
[12] J. Rose, H. Ralston, J.G. Gamble, Energetics of walking, in: J. Rose, J. Gamble [27] R.E. Cowan, M.S. Nash, J.L. Collinger, A.M. Koontz, M.L. Boninger, Impact of
(Eds.), Human Walking, third ed., Lippincott Williams & Wilkins, 2006, pp. surface type, wheelchair weight, and axle position on wheelchair propulsion by
77–102, https://www.worldcat.org/title/human-walking/oclc/474543858. novice older adults, Archives of Physical Medicine and Rehabilitation 90 (7)
[13] A.E. Minetti, C. Moia, G.S. Roi, D. Susta, G. Ferretti, Energy cost of walking and (2009) 1076–1083, http://dx.doi.org/10.1016/j.apmr.2008.10.034.
running at extreme uphill and downhill slopes, Journal of Applied Physiology [28] G.A. Wolfe, R. Waters, H.J. Hislop, Influence of floor surface on the energy
93 (3) (2002) 1039–1046, http://dx.doi.org/10.1152/japplphysiol.01177.2001. cost of wheelchair propulsion, Physical Therapy 57 (9) (1977) 1022–1027,
[14] P.C. Dixon, D.J. Pearsall, Gait dynamics on a cross-slope walking surface, Journal http://dx.doi.org/10.1093/ptj/57.9.1022.
of Applied Biomechanics 26 (1) (2010) 17–25, http://dx.doi.org/10.1123/jab.26. [29] E.G. Gonzalez, P.J. Corcoran, Chapter 16 - energy expenditure during am-
1.17. bulation, in: J.A. Downey, S.J. Myers, E.G. Gonzalez, J.S. Lieberman (Eds.),
[15] R.A. Cooper, E.E. Teodorski, M.L. Sporner, D.M. Collins, Manual wheelchair The Physiological Basis of Rehabilitation Medicine, second ed., Butterworth-
propulsion over cross-sloped surfaces: A literature review, Assistive Technology Heinemann, 1994, pp. 413–446, http://dx.doi.org/10.1016/B978-1-4831-7818-
23 (1) (2011) 42–51, http://dx.doi.org/10.1080/10400435.2010.541408. 9.50022-8.
15
[30] C.E. Brubaker, C.A. McLaurin, I.S. McClay, Effects of side slope on wheelchair [53] X. Li, C. Claramunt, C. Ray, A grid graph-based model for the analysis of
performance, Journal of Rehabilitation Research and Development 23 (2) (1986) 2D indoor spaces, Computers, Environment and Urban Systems 34 (6) (2010)
55–57, https://pubmed.ncbi.nlm.nih.gov/3723426/. 532–540, http://dx.doi.org/10.1016/j.compenvurbsys.2010.07.006.
[31] D. Friedland, I. Brunton, J. Potts, Falls and traumatic brain injury in adults [54] D. Nagy, L. Villaggi, J. Stoddart, D. Benjamin, The buzz metric: A graph-based
under the age of sixty, Journal of Community Health 39 (1) (2014) 148–150, method for quantifying productive congestion in generative space planning for
http://dx.doi.org/10.1007/s10900-013-9752-3. architecture, Technology| Architecture+ Design 1 (2) (2017) 186–195, http:
[32] R. Riener, M. Rabuffetti, C. Frigo, Stair ascent and descent at different incli- //dx.doi.org/10.1080/24751448.2017.1354617.
nations, Gait & Posture 15 (1) (2002) 32–44, http://dx.doi.org/10.1016/S0966- [55] B. Doherty, D. Rumery, B. Barnes, B. Zhou, A spatial query & analysis tool
6362(01)00162-X. for architects, in: Proceedings of the 2012 Symposium on Simulation for
[33] K.A. Hamel, N. Okita, J.S. Higginson, P.R. Cavanagh, Foot clearance during stair Architecture and Urban Design, SimAUD ’12, Society for Computer Simulation
descent: effects of age and illumination, Gait & Posture 21 (2) (2005) 135–140, International, San Diego, CA, USA, 2012, pp. 1–8, http://dx.doi.org/10.5555/
http://dx.doi.org/10.1016/j.gaitpost.2004.01.006. 2339453.2339457.
[34] M. Johnson, B. Shrewsbury, S. Bertrand, T. Wu, D. Duran, M. Floyd, P. Abeles, D. [56] K. Ninomiya, M. Kapadia, A. Shoulson, F. Garcia, N. Badler, Planning approaches
Stephen, N. Mertins, A. Lesman, J. Carff, W. Rifenburgh, P. Kaveti, W. Straatman, to constraint-aware navigation in dynamic environments, Computer Animation
J. Smith, M. Griffioen, B. Layton, T. Boer, T. Koolen, P. Neuhaus, J. Pratt, Team and Virtual Worlds 26 (2) (2015) 119–139, http://dx.doi.org/10.1002/cav.1622.
[57] M. Bielik, S. Schneider, R. König, Parametric urban patterns: Exploring and
IHMC’s lessons learned from the DARPA robotics challenge trials, Journal of
integrating graph-based spatial properties in parametric urban modelling,
Field Robotics 32 (2) (2015) 192–208, http://dx.doi.org/10.1002/rob.21571.
in: Digital Physicality - Proceedings of the 30th ECAADe Conference, vol.
[35] M.G. Dissanayake, P. Newman, S. Clark, H.F. Durrant-Whyte, M. Csorba, A
1, CumInCAD, 2012, pp. 701–708, http://papers.cumincad.org/cgi-bin/works/
solution to the simultaneous localization and map building (SLAM) problem,
paper/ecaade2012_057.
IEEE Transactions on Robotics and Automation 17 (3) (2001) 229–241, http:
[58] J.-k. Lee, C.M. Eastman, J. Lee, M. Kannala, Y.-s. Jeong, Computing walking
//dx.doi.org/10.1109/70.938381.
distances within buildings using the universal circulation network, Environment
[36] S.S. Ge, Q. Zhang, A.T. Abraham, B. Rebsamen, Simultaneous path planning and
and Planning B: Planning and Design 37 (4) (2010) 628–645, http://dx.doi.org/
topological mapping (SP2ATM) for environment exploration and goal oriented
10.1068/b35124.
navigation, Robotics and Autonomous Systems 59 (3–4) (2011) 228–242, http:
[59] W.Y. Lin, P.H. Lin, Intelligent generation of indoor topology (i-GIT) for human
//dx.doi.org/10.1016/j.robot.2010.12.003.
indoor pathfinding based on IFC models and 3D GIS technology, Automation
[37] J.-S. Gutmann, M. Fukuchi, M. Fujita, A floor and obstacle height map for 3D
in Construction 94 (2018) 340–359, http://dx.doi.org/10.1016/j.autcon.2018.07.
navigation of a humanoid robot, in: Proceedings of the 2005 IEEE International
016.
Conference on Robotics and Automation, IEEE, 2005, pp. 1066–1071, http: [60] S. Taneja, B. Akinci, J.H. Garrett Jr., L. Soibelman, Algorithms for automated
//dx.doi.org/10.1109/ROBOT.2005.1570257. generation of navigation models from building information models to support
[38] S. Pütz, T. Wiemann, J. Sprickerhof, J. Hertzberg, 3D Navigation mesh generation indoor map-matching, Automation in Construction 61 (2016) 24–41, http://dx.
for path planning in uneven terrain, IFAC-PapersOnLine 49 (15) (2016) 212–217, doi.org/10.1016/j.autcon.2015.09.010.
http://dx.doi.org/10.1016/j.ifacol.2016.07.734. [61] P. Boguslawski, L. Mahdjoubi, V. Zverovich, F. Fadli, Automated construction of
[39] M. Kallmann, M. Kapadia, Navigation meshes and real-time dynamic planning variable density navigable networks in a 3D indoor environment for emergency
for virtual worlds, in: ACM SIGGRAPH 2014 Courses, SIGGRAPH ’14, Association response, Automation in Construction 72 (2016) 115–128, http://dx.doi.org/10.
for Computing Machinery, New York, NY, USA, 2014, pp. 498–578, http://dx. 1016/j.autcon.2016.08.041.
doi.org/10.1145/2614028.2615399. [62] C. Saona-Vazquez, I. Navazo, P. Brunet, The visibility octree: a data structure
[40] T. Narahara, Self-Organizing Computation a Framework for Generative Ap- for 3D navigation, Computers & Graphics 23 (5) (1999) 635–643, http://dx.doi.
proaches to Architectural Design (Ph.D. thesis), Harvard University, USA, 2010, org/10.1016/S0097-8493(99)00087-4.
pp. 1–318, https://dl.acm.org/doi/book/10.5555/2395396. [63] P. Nourian, Configraphics: Graph Theoretical Methods for Design and Anal-
[41] R. Oliva, N. Pelechano, NEOGEN: Near optimal generator of navigation meshes ysis of Spatial Configurations (Ph.D. thesis), Delft University of Technology,
for 3D multi-layered environments, Computers & Graphics 37 (5) (2013) 2016, pp. 1–348, https://books.bk.tudelft.nl/index.php/press/catalog/book/isbn.
403–412, http://dx.doi.org/10.1016/j.cag.2013.03.004. 9789461867209.
[42] B. Chazelle, Triangulating a simple polygon in linear time, Discrete & [64] J. Roles, H. ElAarag, A Smoothest Path algorithm and its visualization tool, in:
Computational Geometry 6 (3) (1991) 485–524, http://dx.doi.org/10.1007/ 2013 Proceedings of IEEE Southeastcon, 2013, pp. 1–6, http://dx.doi.org/10.
BF02574703. 1109/SECON.2013.6567453.
[43] W. Van Toll, A.F. Cook, R. Geraerts, Navigation meshes for realistic multi-layered [65] L. Liu, R.C.-W. Wong, Finding shortest path on land surface, in: Proceedings
environments, in: 2011 IEEE/RSJ International Conference on Intelligent Robots of the 2011 ACM SIGMOD International Conference on Management of Data,
and Systems, IEEE, 2011, pp. 3526–3532, http://dx.doi.org/10.1109/IROS.2011. SIGMOD ’11, ACM, New York, NY, USA, 2011, pp. 433–444, http://dx.doi.org/
6094790. 10.1145/1989323.1989369.
[44] J. Pettré, J.-P. Laumond, D. Thalmann, A navigation graph for real-time [66] W. Yuan, M. Schneider, Supporting 3D route planning in indoor space based
crowd animation on multilayered and uneven terrain, in: First Interna- on the LEGO representation, in: Proceedings of the 2nd ACM SIGSPATIAL
tional Workshop on Crowd Simulation (V-CROWDS), vol. 43(44), Pergamon International Workshop on Indoor Spatial Awareness, 2010, pp. 16–23, http:
Press, New York, 2005, pp. 194–202, https://pdfs.semanticscholar.org/7d1a/ //dx.doi.org/10.1145/1865885.1865890.
8c164f7690f43c82de53a041c1a165821af5.pdf. [67] T. Narahara, The Space Re-Actor: Walking a Synthetic Man Through Architectural
[45] F. Lamarche, TopoPlan: a topological path planner for real time human naviga- Space, Massachusetts Institute of Technology, 2007, pp. 1–85, http://hdl.handle.
tion under floor and ceiling constraints, in: Computer Graphics Forum, vol. 28(2), net/1721.1/39255.
[68] A. Rahman, A.K. Mahmood, Agent-based simulation using prometheus method-
Wiley Online Library, 2009, pp. 649–658, http://dx.doi.org/10.1111/j.1467-
ology in evacuation planning, in: 2008 International Symposium on Information
8659.2009.01405.x.
Technology, vol. 3, IEEE, 2008, pp. 1–8, http://dx.doi.org/10.1109/ITSIM.2008.
[46] Unity User Manual 2019.1, Building height mesh for accurate character place-
4632061.
ment, 2019, (online), https://docs.unity3d.com/Manual/nav-HeightMesh.html,
[69] S. Gwynne, E.R. Galea, M. Owen, P.J. Lawrence, L. Filippidis, A review of
(Accessed 19 July 2019).
the methodologies used in the computer simulation of evacuation from the
[47] R. Geraerts, Planning short paths with clearance using explicit corridors, in:
built environment, Building and Environment 34 (6) (1999) 741–749, http:
2010 IEEE International Conference on Robotics and Automation, IEEE, 2010,
//dx.doi.org/10.1016/S0360-1323(98)00057-2.
pp. 1997–2004, http://dx.doi.org/10.1109/ROBOT.2010.5509263.
[70] X. Pan, C.S. Han, K. Dauber, K.H. Law, A multi-agent based framework for the
[48] R. Geraerts, M.H. Overmars, Enhancing corridor maps for real-time path planning
simulation of human and social behaviors during emergency evacuations, AI &
in virtual environments, in: Computer Animation and Social Agents, 2008, pp.
Society 22 (2) (2007) 113—132, http://dx.doi.org/10.1007/s00146-007-0126-1.
64–71, https://dspace.library.uu.nl/handle/1874/40529. [71] D. Schaumann, N.P. Pilosof, H. Sopher, J. Yahav, Y.E. Kalay, Simulating
[49] J.L. Vermeulen, A. Hillebrand, R. Geraerts, Annotating traversable gaps in multi-agent narratives for pre-occupancy evaluation of architectural designs, Au-
walkable environments, in: 2018 IEEE International Conference on Robotics and tomation in Construction 106 (2019) 1–19, http://dx.doi.org/10.1016/j.autcon.
Automation, ICRA, IEEE, 2018, pp. 3045–3052, http://dx.doi.org/10.1109/ICRA. 2019.102896.
2018.8461152. [72] E.V. Lang, A better patient experience through better communication, Journal of
[50] J.A. Jakubiec, C.F. Reinhart, DIVA 2.0: Integrating daylight and thermal simula- Radiology Nursing 31 (4) (2012) 114–119, http://dx.doi.org/10.1016/j.jradnu.
tions using Rhinoceros 3D, Daysim and EnergyPlus, in: Proceedings of Building 2012.08.001.
Simulation, vol. 12, 2011, pp. 2202–2209. [73] M. Nakada, T. Zhou, H. Chen, T. Weiss, D. Terzopoulos, Deep learning of
[51] T. Lozano-Pérez, M.A. Wesley, An algorithm for planning collision-free paths biomimetic sensorimotor control for biomechanical human animation, ACM
among polyhedral obstacles, Communications of the ACM 22 (10) (1979) Transactions on Graphics 37 (4) (2018) http://dx.doi.org/10.1145/3197517.
560–570, http://dx.doi.org/10.1145/359156.359164. 3201305.
[52] A. Turner, M. Doxa, D. O’sullivan, A. Penn, From isovists to visibility graphs: a [74] G. Suter, Structure and spatial consistency of network-based space layouts
methodology for the analysis of architectural space, Environment and Planning B: for building and product design, Computer-Aided Design 45 (8–9) (2013)
Planning and Design 28 (1) (2001) 103–121, http://dx.doi.org/10.1068/b2684. 1108—1127, http://dx.doi.org/10.1016/j.cad.2013.04.004.
16
[75] M.L. Benedikt, To take hold of space: Isovists and isovist fields, Environment [80] E.W. Dijkstra, A note on two problems in connexion with graphs, Numerische
and Planning B: Planning and Design 6 (1) (1979) 47–65, http://dx.doi.org/10. Mathematik 1 (1) (1959) 269–271, http://dx.doi.org/10.1007/BF01386390.
1068/b060047. [81] M. Schwartz, S. Das, Interpretting non-flat surfaces for walkability analysis, in:
[76] T. Varoudis, S. Psarra, Beyond two dimensions: Architecture through three- Proceedings of the Symposium on Simulation for Architecture and Urban Design,
dimensional visibility graph analysis, The Journal of Space Syntax 5 (1) (2014) SIMAUD ’19, Society for Computer Simulation International, 2019, pp. 287–294,
91–108, http://joss.bartlett.ucl.ac.uk/journal/index.php/joss/article/view/200. https://dl.acm.org/doi/10.5555/3390098.3390128.
[77] E. Fuchkina, Pedestrian movement graph analysis, in: Arbeitspapiere Informatik
[82] S.K. Ghosh, P.P. Goswami, Unsolved problems in visibility graphs of points,
in der Architektur (Working Papers Computer Science In Architecture), vol.
segments, and polygons, ACM Computing Surveys 46 (2) (2013) 22:1–22:29,
16, Bauhaus-University Weimar, Weimar, 2016, pp. 1–18, http://dx.doi.org/10.
http://dx.doi.org/10.1145/2543581.2543589.
25643/bauhaus-universitaet.2738.
[78] J.-K. Lee, J. Shin, Y. Lee, Circulation analysis of design alternatives for elderly [83] ESRI, CityEngine, 2020, (online), https://www.esri.com/en-us/arcgis/products/
housing unit allocation using building information modelling-enabled indoor esri-cityengine/overview, (Accessed 7 January 2020).
walkability index, Indoor and Built Environment (2018) 355–371, http://dx.doi. [84] L.-F. Yu, S.-K. Yeung, C.-K. Tang, D. Terzopoulos, T.F. Chan, S.J. Osher, Make it
org/10.1177/1420326X18763892. home: Automatic optimization of furniture arrangement, ACM Transactions on
[79] M. Schwartz, A. Azeez, K. Patel, Human task and disability based automated Graphics 30 (4) (2011) http://dx.doi.org/10.1145/2010324.1964981.
evaluation of space and design in CAD, in: Proceedings of the Symposium on [85] T. Weiss, M. Nakada, D. Terzopoulos, Automated layout synthesis and visual-
Simulation for Architecture and Urban Design, SIMAUD ’18, Society for Computer ization from images of interior or exterior spaces, in: 2017 IEEE Conference on
Simulation International, 2018, pp. 181–188, http://dl.acm.org/citation.cfm?id= Computer Vision and Pattern Recognition Workshops, CVPRW, 2017, pp. 41–47,
3289750.3289768. http://dx.doi.org/10.1109/CVPRW.2017.12.
17

Automation in Construction: Mathew Schwartz

Uploaded by

Automation in Construction: Mathew Schwartz

Uploaded by

Automation in Construction 127 (2021) 103557

Contents lists available at ScienceDirect

Human centric accessibility graph for environment analysis

ARTICLE INFO ABSTRACT

E-mail addresses: cadop@njit.edu, cadop@umich.edu.

1. A novel ray-casting based method for creating a grid-based

The paper is organized as follows; Section 2 presents a review of

Fig. 6. Representation of the environments into a grid of nodes. Illustration showing

3.1.1. Surface definition

metric, such as visibility, information about the access to a node is

Algorithm 1: Build digraph of nodes and edges

In Fig. 10, a simple overview of the algorithm to build the

queries of the relationship between nodes (such as defining edge cost)

𝛹𝑐 (𝑒𝑖𝑗 ) = Cross-slope, 𝛹𝑑 (𝑒𝑖𝑗 ) = Metric Distance, 𝛹𝑠 (𝑒𝑖𝑗 ) = Slope, 𝛹𝑒 (𝑒𝑖𝑗 )

You might also like