Generation of Dimensioned Floor Plans for a Given Boundary Layout

Journal of Soft Computing in Civil Engineering 8-3 (2024) 87-106
How to cite this article: Pinki, P., Shekhawat, K., Goswami, D., Jain, U. Generation of dimensioned floor plans for a given
boundary layout. J Soft Comput Civ Eng 2024;8(3):87–106. https://doi.org/10.22115/scce.2023.377613.1583
2588-2872/ © 2024 The Authors. Published by Pouyan Press.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Contents lists available at SCCE
Journal of Soft Computing in Civil Engineering
Journal homepage: www.jsoftcivil.com
Generation of Dimensioned Floor Plans for a Given Boundary
Layout
Pinki Pinki 1,*
; Krishnendra Shekhawat 2
; Dipam Goswami 3
; Ujjwal Jain 3
1. Post Doctoral Fellow, Birla Institute of Technology and Science, Pilani, India
2. Associate Professor, Faculty of Mathematics, Birla Institute of Technology and Science, Pilani, India
3. Undergraduate Student, Birla Institute of Technology and Science, Pilani, India
Corresponding author: 456pinkiyadav@gmail.com
https://doi.org/10.22115/SCCE.2023.377613.1583
ARTICLE INFO ABSTRACT
Article history:
Received: 27 December 2022
Revised: 08 August 2023
Accepted: 31 December 2023
In literature, the generation of floor plans has mainly been
confined to dimensionless floor plans with rectangular
boundaries having no open spaces within the floor plan. In
this paper, the user is allowed to construct a dimensioned
boundary using slant, horizontal and vertical line segments,
where dimensioned open spaces can be drawn within the
boundary layout. Once the boundary is finalized, it can be
partitioned into dimensioned blocks using vertical and
horizontal dissections. Each block can be further partitioned
into dimensioned rooms which results in a dimensioned floor
plan F for the given boundary layout. The dissection method
employed is based on the slicing tree approach, which results
in floor plans that are amenable to slicing and consist of non-
rectangular rooms, offering the potential for open spaces
within floor plan F. As a preliminary step towards
automation, we have developed an interactive user interface
for generating dissected dimensioned floor plans.
Keywords:
Rooms;
Non-rectangular floor plan;
Slicing tree;
Algorithm;
Dissection.

88 P. Pinki et al./ Journal of Soft Computing in Civil Engineering 8-3 (2024) 87-106
1. Introduction
A floor plan (FP) is defined as a closed polygon divided by straight-line segments into a finite
number of non-overlapping polygons called rooms. The shapes of rooms and floor plan boundary
can be used to classify FPs. A FP is called a rectangular floor plan (RFP) if it has a rectangular
boundary and consists of rectangular rooms. A RFP has been classified as slicing FPs and non-
slicing FPs. A slicing FP is a rectangular dissection obtained by dissecting a rectangle
horizontally or vertically into two smaller rectangles in a recursive manner (Figure 1a). An
ordered tree, also known as a slicing tree [1], is the earliest representation used by researchers to
represent slicing FPs in the literature [2]. Each internal node represents a vertical or a horizontal
merge operation on the two descendants, while each leaf of the slicing tree corresponds to a
block. A FP is called an orthogonal floor plan (OFP) if its boundary is rectangular, and it has
rectangular and rectilinear rooms (at least one rectilinear). A non-rectangular floor plan (NRFP)
has a rectilinear or slanted boundary. A rectilinear floor plan is a specific case of NRFP having a
rectilinear boundary and rectangular or rectilinear rooms (Figure 1).
(a) Slicing floor plan (b) Non-slicing
floor plan
(c) Orthogonal
floor plan
(d) Rectilinear floor
plan
(e) Non-rectangular
floor plan
Fig. 1. Types of Floor Plans.
In recent years, researchers have focused on the automated generation of floor plans. The
generation of a floor plan is an important step in the design process and it has different problem
perspectives. From an architectural perspective, it is a design problem treated as a preliminary
layout which can be further modified and adjusted by architects. From a mathematical
perspective, it is a polygonal representation of vertices which satisfies the adjacency
requirements of a given graph. From a VLSI perspective, it is an optimization problem where it
is required to optimize the chip area while structuring its design. Clearly, the problem of
generating floor plans is of an interdisciplinary nature, which reflects the growing interest of
researchers in the automated generation of floor plans. Therefore, there exist various approaches
to automate the generation of floor plans which are based on graph theory, optimization,
dissection, shape grammar, neural networks, machine learning and artificial intelligence, etc. At
the same time, none of the approaches can be used to generate floor plans for a given boundary
layout, in particular non-rectangular.
In this paper, we provide an interactive tool for the generation of dimensioned floor plans by
dissecting a given plot area. Initially, a user-specific non-rectangular boundary layout has been
created from a given rectangular plot. The user has the autonomy to determine the dimensions of
the boundary which is made up of slant, horizontal and vertical line segments as shown in Figure

P. Pinki et al./ Journal of Soft Computing in Civil Engineering 8-3 (2024) 87-106 89
2a. Within the boundary, the user has a choice to insert dimensioned open spaces by selecting the
interior points. After finalising the boundary, the layout can be partitioned into dimensioned
blocks using vertical and horizontal dissections as shown in Figures 2b and 2d.
(a) (b) (c)
(d) (e) (f)
(g)
Fig. 2. Customised generation of a dimensioned floor plan using dissection method (a) A given layout
with dimensioned non-rectangular boundary, (b) and (d) Types of dissection (horizontal or vertical
dissection) required to create rectilinear/rectangular rooms, (c) and (e) Input dimension based on the
dissection type for creating rectilinear/rectangular rooms, (f) Selecting a block to make it a room, (g)
Required floor plan.

After slicing a block horizontally, let the newly formed blocks be A and B, where block A is the
upper block and block B is the lower block. For horizontal slicing, the user has to input the
height of the upper block A, since its width must be equal to the width of the parent block to
ensure the generation of a FP as shown in Figures 2d and 2e. Similarly, after slicing a block
vertically, let the newly formed blocks be C and D, where block C is the left block and block D is
the right block. For vertical slicing, the user has to input the width of the left block C as shown in
Figures 2b and 2c. For each block, the user has a choice to either consider it as a room or further
partition it into dimensioned rooms as per requirements, which results in a dimensioned floor
plan for the given boundary layout (refer to Figure 2 for the step-by-step explanation of the
construction of dimensioned FP from a non-rectangular boundary). The proposed algorithm is
implemented using Python programming language and it can be used for customized generation
of dimensioned RFPs, OFPs, and NRFPs, with or without open spaces. Our work is the first step
towards automating the generation of dimensioned floor plans for given non-rectangular
boundary layouts, where adjacency relations would be incorporated in future work.
The rest of the paper is structured as follows. Section 2 examines the related work. Section 3
explains the concepts and definitions used in our approach. Section 4 presents the construction of
dimensioned FPs with or without open spaces and provides an explanatory example. This section
also provides the pseudocode for the proposed algorithm. Section 5 describes the results and
discussion part along with the regeneration of some existing FPs. Section 6 concludes the paper
and proposes future scope.
2. History of floor plans
The term “floor plan generation” has piqued the interest of researchers due to its broad range of
applications in different fields such as architecture, VLSI, etc. Due to the growing attention, the
problem has been intensively studied and different methodologies for the generation of FPs have
been used in literature.
2.1. Generation of floor plans
Graph-theoretic approaches: The problem of generating FPs using graph theory has been
initiated in 1960s by Levin [3]. He proposed a representation for an adjacency graph in the form
of a FP. In 1970s, the problem has been extensively studied [4–8]. The problem of representing a
graph in the form of a FP is mathematically challenging because not all graphs can have a FP (for
example, non-planar graphs). Hence, it has been addressed on the basis of classes of graphs (i.e.,
outerplanar graphs1
, maximal planar graphs (MPGs)2
, planar triangulated graphs (PTGs)3
, 1-
connected PTGs4
, bi-connected PTGs5
etc.) and types of FPs (RFP, OFP, rectilinear FP). In 1985,
1
An outerplanar graph is a graph that can be embedded in the plane with all of its vertices lying on the exterior face.
2
A maximal planar graph (MPG) is a planar graph whose every face is a triangle and adding an edge to it would destroy its
planarity.
3
A planar triangulated graph (PTG) is a planar graph with no simple cycles of length 4 or more.
4
A PTG with cut vertices.
5
A PTG with no cut vertices.

Koźmiński and Kinnen gave a necessary and suﬀicient condition for the existence of a RFP for a
given PTPG6
. There exist many other approaches to generate RFPs, i.e., coloring technique [9],
regular edge labelling [10], slicing tree [11,12], constraint satisfaction [13] and Monte-Carlo Tree
search-based reinforcement learning [14], etc. Other works on the generation of RFPs can be
found at [10,15–23].
In 1993, Yeap and Sarrafzadeh [24] discovered that there exists a graph for which RFPs do not
exist. In the later years, a lot of research has been conducted on OFPs (rectilinear
representations) for MPGs using different approaches such as face area concept, canonical
labelling, duality, etc. [25–34]. Recently, the work presented by Shekhawat et al. [35]
demonstrated the utilization of graph algorithms for the generation of OFPs with room shapes
tailored according to user specifications.
Due to the applications of FPs in VLSI circuits and architectural layout arrangements, FPs are
not limited to rectangular boundaries only. In 1982, Baybars et al. [36] proposed the generation
of FPs with rectilinear boundaries and circulation spaces for a given MPG using the concept of
pseudo-geometric dual. In 1992, Nummenmaa [37] gave an algorithm to construct rectilinear FPs
for MPGs using the canonical representation of planar graphs. In 2020, Wang et al. [38]
proposed FP generation with rectilinear boundaries and rooms. In 2022, Sumit et al. [39]
presented the automated generation of dimensioned and dimensionless FPs. They proposed a
software GPLAN which uses graphic theoretic algorithms along with linear optimization
techniques. In a recent study, Raveena et al. [40] introduced the necessary and sufficient
conditions for the existence of non-trivial L-shaped FPs corresponding to PTPGs and gave an
algorithm for its construction, if it exists.
Other approaches: In the last few decades, researchers have explored different approaches for the
automatic generation of FPs. In 2010, Merrell et al. [41] introduced supervised learning for
architectural design and proposed the automated generation of FPs using Bayesian networks. In
2013, [42] devised an enhanced hybrid evolutionary computation scheme for generating FPs that
combines an Evolutionary Strategy (ES) and a Stochastic Hill Climbing (SHC) technique. In
2018, Wu et al. [43] presented the generation of dimensioned FPs using mixed quadratic
programming. There are many other approaches for the generation of FPs, i.e., neural networks
[44], generic optimization [45,46], artificial intelligence [47], data-driven techniques [48,49],
constrained optimization [50], generative adversarial layout refinement network [51], integer
linear programming [52], Shape grammar [53–55] etc. For more work on the generation of
rectilinear FPs corresponding to the user-specified design requirements, refer to [38,44,48,56]. In
2023, Wang et al. [57] proposed a framework for automated building layout generation,
leveraging the power of deep learning and graph algorithms. On the other hand, Weber et al. [58]
conducted a recent study that critically reviewed various methods and their applications for
automating architectural FP design.
6
A properly triangulated planar graph (PTPG) is a connected planar graph where all interior faces have length 3, internal
vertices have degree ≥ 4 and non-face cycle with length ≥ 4.

Dimensioned FPs: Ideas for the work on dimensioned FPs were proposed in 1970s by Cousin
[4]. In 1982, Roth et al. [9] proposed dimensioned FPs for an adjacency graph using the PERT
algorithm. In 1987, Rinsma [16] presented the dimensioned RFPs and OFPs for a given tree.
Many researchers used weighted graphs as an input to generate dimensioned FPs [19,30,59]. For
recent work on dimensioned FPs, refer to [38,60,61].
2.2 Significance of floor plans with non-rectangular boundary layout and proposed work
In modern times, design cannot remain restricted to rectangular or rectilinear boundaries. It can
be observed that for complex building structures like schools, hospitals and oﬀice buildings etc.,
we may require layouts with non-rectangular boundary having open spaces within the layout.
Hence, this work is a primary step towards the automated generation of floor plans with non-
rectangular boundaries.
By reviewing the existing work on FPs, it can be observed that the construction of FPs is mainly
restricted to FPs with rectangular or rectilinear boundaries. Also, the work on the rectilinear
boundary is restricted to specific shapes like L-shape, T-shape, staircase, etc. (refer to
[33,38,43,44,48,56]) with no internal spaces (atrium) within the FP. Also, there is no work on
FPs which consider slanted line segments for a boundary layout. In this work, we propose a
graphical user interface for generating the initial architectural dimensioned layout of a FP with
no restriction on the boundary, i.e., it is feasible to consider a dimensioned plot with a
rectangular or non-rectangular boundary or slanted boundary. Also, it is possible to add
dimensioned atrium within the plot area. Using the dissection method, the required plot would be
partitioned into the required number of dimensioned rooms to have a required floor plan.
3. Preliminaries
Here, we cover basic definitions and their notations which are used frequently in this paper.
These terms have been previously mentioned in literature [62] and are explained here for a better
understanding.
Definition 1 [FP]: A floor plan (FP) is a closed polygon partitioned into a finite number of non-
overlapping polygons by straight/slant line segments. A room in a FP is a polygon that cannot be
further subdivided. A FP with rectangular boundary and rectangular rooms is called a rectangular
floor plan (RFP). RFPs are mainly categorised into the following two types:
i. Slicing FP
ii. Non-slicing FP
Definition 2 [SFP]: Slicing floor plan (SFP) is obtained by finitely repeated bisection of a larger
rectangle until each part is a single rectangle (Figure 1a). It can be represented by a rooted binary
tree known as a slicing tree (ST). Leaves of a ST represent rooms of a FP and its internal vertices
represent cut types indicated by H (for horizontal division) and V (for vertical division). For
example, in Figure 3, a SFP and its corresponding ST are given. There can be more than one ST
corresponding to a FP (Figure 4).

Definition 3 [NSFP]: A Non-slicing floor plan (NSFP) is a more general FP. It cannot be
bisected and hence cannot be represented by ST.
Fig. 3. A slicing floor plan and its corresponding slicing tree.
Fig. 4. Different representations of a Slicing Floor Plan given in Figure 1(a).
Definition 4 [Block]: A block in a FP is a polygon which has either a vertical or a horizontal
division that divides the block into two smaller blocks, i.e., child blocks. If a block has a vertical
division, then its child block has the same height as that of the parent block. Similarly, if a block
has a horizontal division, then its child block has the same width as that of the parent block. A FP
is a block in itself.
Notations:
FP: floor plan, RFP: rectangular FP, OFP: orthogonal FP, NRFP: non-rectangular FP, SFP:
slicing FP, NSFP: non-slicing FP, MPG: maximal planar graph, PTG: planar triangulated graph,
PTPG: properly triangulated planar graph.
4. Construction of dimensioned FPs with or without open spaces
This section describes the working of the proposed algorithm which is implemented using
Python programming language. The proposed algorithm is used for the customized generation of

a dimensioned FP starting from a given non-rectangular dimensioned layout and gradually
dissecting it horizontally and vertically. Its steps are as follows:
Step 1: Creating a rectangular plot
To create a rectangular plot, the user needs to enter its width and height, which generates a
rectangular plan on the screen with the required width and height (Figure 5).
Fig. 5. Entering dimensions of the required rectangular plan and corresponding obtained plan.
Step 2: Forming a dimensioned boundary layout
Step 2.1: Selection of boundary points
To form a FP boundary, the user needs to select points on the boundary of the rectangular plot
with two choices resulting in slanted line segments or rectilinear line segments (Figure 6).
Fig. 6. Selection of boundary points.
(a) (b)
Fig. 7. (a) Selection of interior points, (b) Finalised non-rectangular dimensioned boundary layout with
atrium.

Step 2.2: Selection of interior points
Selecting the points inside the plan creates open spaces (Figure 7). After finalizing the boundary
points and interior points, a dimensioned boundary layout is obtained as shown in Figure 7(b).
Step 3: Constructing a FP using dissection
To construct a required FP, the user can sequentially dissect the plan obtained in Step 2, either
horizontally or vertically, by right-clicking on a single block at a time. After each dissection, a
block area gets partitioned into two sub-blocks. If the user does not want to further partition a
block, then they have the option to designate the block as a room.
Step 3.1: Vertical and horizontal dissection
Each block in a FP is considered as a node in the slicing tree representation of the FP. Every
block has the following attributes: parent, left_child, right_child, height, width, slice_type, d1,
d2, d3 and d4. The attributes parent, left_child, and right_child point to another block in the FP,
height and width store dimensions of a particular block, slice_type stores the type of slicing on
its parent block and d1, d2, d3, d4 stores the top-left corner x-coordinate, top-left corner y-
coordinate, bottom right corner x-coordinate, bottom right corner y-coordinate of the block
respectively. The attributes for all the blocks are stored independently for each of them. On
dissecting a block, it gets connected to its left child and right child representing the two blocks
formed as a result of dissection.
Step 3.2: Block dimensioning
For horizontal slicing, the user has to input the height of the upper block of dissection since its
width equals the width of the parent block to ensure the generation of a FP. We obtain the height
of the lower block by subtracting the height of the upper block from the height of the parent
block. Similarly, for vertical slicing, the user has to input the width of the left block of dissection.
Two new blocks formed by dissection have the same parent. The left_child and right_child
properties of the parent block now point to the new blocks. The height and width of the new
blocks are set to the input dimensions. The d1, d2, d3, and d4 values of these blocks are set
according to their positions in the FP. These values increase on moving downward and rightward
(refer to Figure 8 for a better illustration of the positioning of blocks in a FP).
Step 3.3: Formation of rooms
If the user chooses a block to make it a room, then this block is considered as a leaf of the slicing
tree (refer to Figure 9(a) for an illustration of the step where the choice is given for making the
block as a room). A room is represented by a shaded rectangle and labeled with a room number
where the dimensions of the room are displayed below the room name (refer to Figure 9(b)
which is a dimensioned FP and for the illustration of all the steps for the transformation of a plot
into a dimensioned FP, refer Figures 5 to 9).

(a) (b)
(c) (d)
(e) (f)
Fig. 8. (a) Choice for the type of dissection (b) After opting for horizontal dissection, the dimension of the
block needs to be given as input, which would show an error if horizontal dissection is not possible for
the provided dimensions, (c) Horizontal dissection of the block satisfying dimensions as given in (b), (d)
Again providing the choices for the type of dissection,(e) After opting for vertical dissection, dimension
of the block needs to be given which would show an error if the vertical dissection is not possible for
provided dimensions, (f) Vertical dissection of the block satisfying dimensions given in (e). (a) Selection
of interior points, (b) Finalised non-rectangular dimensioned boundary layout with an atrium.

4.1. Pseudo code for the construction of dimensioned FPs
In this section, we present pseudo code for the generation of FPs which is implemented using
Python programming language with the help of its GUI library Tkinter.
Algorithm 1. FP
Input: A rectangular plot and a non-rectangular boundary layout A
Output: A dimensioned floor plan
Data: Dimension of a floor plan and its dissected components
1 Create a rectangular plot with given dimensions (Width of plot > 0, Height of plot > 0)
2 Finalise the type of FP
(a) select external points for having boundary with rectilinear and slant line segments
(b) select points inside the plot for creating atriums
3 while (all blocks are not rooms) do
4 Choose the block of a plot to dissect
5 Choose vertical or horizontal dissection for the chosen block of a plot
6 if Slice type is vertical then
7 Height for left sub-block = Height of block
8 Width for left sub-block = Input by user
9 if Width of left sub-block < Width of block then
10 Partition the selected block vertically with user input dimensions
11 else
12 Dissection is not feasible with given dimensions
13 Width has to be less than width of the block
14 Enter width for left sub-block
15 else if Slice type is horizontal then
16 Width for upper sub-block = Width of block
17 Height for upper sub-block = Input by user
18 if Height of upper sub-block < Height of block then
19 Partition the selected block horizontally with user input dimensions
20 else
21 Dissection is not feasible with given dimensions
22 Height has to be less than height of the block
23 Enter height for upper sub-block
24 else
25 Assign room name
26 Obtained plan is the desired dimensioned FP
27 Exit

(a) (b)
(c)
Fig. 9. (a) Choice for the type of rooms, (b) Formation of rooms from the dissected blocks, (c) Obtained
dimensioned FP.
5. Results and discussion
In this section, we show some instructive examples that give a quick overview of the proposed
prototype by demonstrating various types of dimensioned FPs that it can generate. This section
also shows how it can be used to regenerate various existing floor plans.
5.1. Regeneration of existing RFP using our prototype:
1. Regeneration of Plan Topology of Villa Badoer
The generated FP maintains the following constraints given in [63] (Figure 10):
(i) The sala ‘R5’ is the largest room.
(ii) The aspect ratio of any room is less than or equal to 2:1 except ‘R7’.

(iii) Rooms surrounding the sala are not identical in shape.
(iv) The ratio of the area of the largest room, the sala to that of the smallest room is 9:1.
(v) No room is smaller than 40 square feet.
(vi) Width of the cells in the underlying grid lying on the axis of symmetry is wider than adjacent
cells.
(vii) No room has dimensions less than 7 feet.
2. Regeneration of Plan Topology of Palladio’s Palazzo Della Torre [58]
Palladio’s Palazzo Della Torre is the fourth building in Andrea Palladio’s Four Books on
Architecture, Book II. Count Giovanni Battista Della Torre of Verona, Italy, commissioned it in
1555 but died before Palazzo could be completed. Here, dimensions are taken based on the
representation of the aspect ratio in the existing plan.
(a)
(b)
Fig. 10. Floor Plan Topology of Villa Badoer7
(a) Existing Floor Plan and its topology [63] (b)
Regenerated Floor Plan.
7
http://www.projekte.kunstgeschichte.uni-muenchen.de/arch_complete_vers/40-ren-barock-
architektur/studieneinheiten/lektion_11/XI_2_10.htm

5.2. Explanatory examples
Using the prototype presented in this paper, different types of FPs can be generated (Figure 12).
Some of them are given as follows:
1. Type 1: NRFP with rectilinear boundary and rectilinear rooms with no open spaces (Figure
12b)
2. Type 2: NRFP with rectilinear boundary, rectilinear rooms and with open spaces (Figure 12c)
3. Type 3: NRFP with slanted boundary and no open spaces (Figure 12a)
4. Type 4: NRFP with slanted boundary and open spaces
5. Type 5: NRFP with the combination of slanted and rectilinear boundary (Figure 12a, e, f)
6. Type 6: RFP (Figure 10)
(a)
(b)
Fig. 11. Floor Plan Topology of Palazzo Della Torre (a) Existing Floor Plan and its topology8
(b)
Regenerated Floor Plan.
8
https://www.ribapix.com/Palazzo-Dalla-Torre-Verona-plan-and-facade_RIBA126891#

(a) (b)
(c) (d)
(e) (f)
Fig. 12. Different types of FPs generated by the prototype.

6. Conclusion and future work
In this paper, we proposed a semi-automatic tool for the generation of dimensioned floor plans
for a given boundary layout. The dimensions of the boundary are chosen by the user which is
made up of slant as well as straight line segments. Within the boundary, there is a choice to insert
dimensioned open spaces. At this stage, the tool could be helpful to architects where they have a
choice to draw the layout boundary and then the boundary can be dissected by first partitioning it
into different dimensioned blocks and further each block is dissected into required rooms. Here,
a dissection is done by simply inputting the width and height of the required block. For each
dissection, there is a choice of vertical as well as horizontal dissection.
The proposed work is the first work considering any dimensioned boundary layout for floor plan
generation. Also, it presents a GUI developed in Python for making the work accessible to the
users. At the same time, to make the tool more useful, we need to automate the process of
dissection by considering certain architectural rules. For example, a bedroom must have a
window and it must be at least 8 feet wide, i.e., we cannot allow the user to make the bedroom as
an interior room. The next challenge will be to consider the adjacency requirements for the
required rooms. Limitations will be discussed in detail in the next section. A brief outline of the
proposed work is as follows:
1. Within existing literature, the floor plans having boundaries made up of slant lines and the
inclusion of atriums have not been explored. This study marks an initial stride toward the
automated generation of dimensioned floor plans within any given arbitrary non-rectangular
boundary. Clearly, the proposed methodology offers a diverse spectrum of choices to users for
space allocation within a given layout, facilitating the creation of non-rectangular dimensioned
floor plans with desired shapes, as depicted in Figure 12.
2. The prototype introduced in this paper employs the dissection method, ensuring that the
generated blocks or rooms adhere to the user's designated dimensional constraints. It also
possesses the ability to generate symmetric dimensioned floor plans (as shown in Figures 10 and
11). Furthermore, it offers a range of alternative layouts, as each block possesses the flexibility to
incorporate spaces in multiple directions, allowing for extensive customization to suit specific
requirements.
The problem of generating dimensioned FPs corresponding to a given adjacency graph with
arbitrary input non-rectangular boundary layout is mathematically very challenging, which we
aim to address in future.
7. Limitations
The study represents an initial stride towards the automated generation of dimensioned floor
plans with non-rectangular boundaries. It exhibits certain limitations that are earmarked for
future refinement.

The prototype necessitates user involvement in determining where dissections should occur to
ensure that the area requirements of individual rooms are satisfied. This entails providing
dimensions for each slicing, thus requiring user input for customization. This constraint could
potentially be mitigated by taking room dimensions as input and subsequently automating the
dissection process.
Furthermore, there are several promising avenues for future research and development,
including:
1. Exploring the generation of floor plans by considering aspect ratios as input rather than
taking dimensions for every block.
2. Investigating the feasibility of generating floor plans by employing a dissection tree as
input which introduces adjacency constraints to the problem.
3. Extending the capabilities of the prototype to generate floor plans with circulations. A
circulation is an interior space of a FP through which people can move. In future, we try to
generate FP with the shortest circulation path spanning all the rooms with one or more
entrances.
4. Qualitative analysis of the obtained floor plans by architects.
Addressing these limitations and pursuing these avenues of exploration has the potential to
significantly enhance the robustness and applicability of our approach.
Funding
This research received no external funding.
Conflicts of interest
The authors declare no conflict of interest.
Authors contribution statement
PP, KS, DG: Conceptualization; PP, DG: Roles/Writing – original draft; PP, DG, UJ: Software,
Visualization; KS: Supervision; PP, KS: Writing – review & editing.
Supplementary material
Regarding the implementation of the proposed work in Python, a video showing the generation
of dimensioned sliceable floor plan layout with rectilinear rooms and non-rectangular
boundaries, with open spaces in the interior is available at
https://www.dropbox.com/scl/fi/ilv1jph3jybw6s2aom0ny/RPF.mp4?rlkey=xumagnscv67xwgi52
1559m253&dl=0.
References
[1] Szepieniec AA, Otten RHJM. The genealogical approach to the layout problem. Proc. seventeenth
Des. Autom. Conf. Des. Autom. - DAC ’80, New York, New York, USA: ACM Press; 1980, p.
535–42. https://doi.org/10.1145/800139.804582.

[2] Otten RHJM. Automatic Floorplan Design. 19th Des. Autom. Conf., IEEE; 1982, p. 261–7.
https://doi.org/10.1109/DAC.1982.1585510.
[3] Levin PH. Use of graphs to decide the optimum layout of buildings. Archit J 1964;7:809–15.
[4] Cousin J. Organisation topologique de l’espace architectural= Topological organization of
architectural spaces. (No Title) 1970.
[5] Grason J. A dual linear graph representation for space-filling location problems of the floor plan
type. Emerg Methods Environ Des Plan 1970.
[6] Steadman P. Graph theoretic representation of architectural arrangement. Archit Res Teach
1973:161–72.
[7] Mitchell WJ. An approach to automated generation of minimum cost dwelling unit planes:
International Technical Cooperation Centre, Tel-Aviv (ITCC) Review, Vol 4, No 3,(1975), pp 116–
139. Comput Des 1977;9:69.
[8] Sauda EJ. Computer program for the generation of dwelling unit floor plans 1975.
[9] Roth J, Wachman A. A Model for Optimal Compact Packing of Convex Cells in an Ortogonal
Network. Technion-IIT, Faculty of Architecture and Town Planning; 1978.
[10] Kant G, He X. Regular edge labeling of 4-connected plane graphs and its applications in graph
drawing problems. Theor Comput Sci 1997;172:175–93. https://doi.org/10.1016/S0304-
3975(95)00257-X.
[11] Del Río-Cidoncha G, Martínez-Palacios J, Iglesias JE. A multidisciplinary model for floorplan
design. Int J Prod Res 2007;45:3457–76. https://doi.org/10.1080/00207540600889550.
[12] Marson F, Musse SR. Automatic Real-Time Generation of Floor Plans Based on Squarified
Treemaps Algorithm. Int J Comput Games Technol 2010;2010:1–10.
https://doi.org/10.1155/2010/624817.
[13] Regateiro F, Bento J, Dias J. Floor plan design using block algebra and constraint satisfaction. Adv
Eng Informatics 2012;26:361–82. https://doi.org/10.1016/j.aei.2012.01.002.
[14] Shi F, Soman RK, Han J, Whyte JK. Addressing adjacency constraints in rectangular floor plans
using Monte-Carlo Tree Search. Autom Constr 2020;115:103187.
https://doi.org/10.1016/j.autcon.2020.103187.
[15] Bhasker J, Sahni S. A linear algorithm to find a rectangular dual of a planar triangulated graph.
Algorithmica 1988;3:247–78. https://doi.org/10.1007/BF01762117.
[16] Rinsma I. Rectangular and Orthogonal Floorplans with Required Room Areas and Tree Adjacency.
Environ Plan B Plan Des 1988;15:111–8. https://doi.org/10.1068/b150111.
[17] Lai Y-T, Leinwand SM. A theory of rectangular dual graphs. Algorithmica 1990;5:467–83.
https://doi.org/10.1007/BF01840399.
[18] He X. On Finding the Rectangular Duals of Planar Triangular Graphs. SIAM J Comput
1993;22:1218–26. https://doi.org/10.1137/0222072.
[19] Eppstein D, Mumford E, Speckmann B, Verbeek K. Area-Universal and Constrained Rectangular
Layouts. SIAM J Comput 2012;41:537–64. https://doi.org/10.1137/110834032.
[20] Shekhawat K, Duarte JP. Introduction to generic rectangular floor plans. Artif Intell Eng Des Anal
Manuf 2018;32:331–50. https://doi.org/10.1017/S0890060417000671.
[21] SHEKHAWAT K, PINKI P. Construction of Architectural Floor Plans for Given Adjacency
Requirements 2020.
[22] Kumar V, Shekhawat K. A transformation algorithm to construct a rectangular floorplan. Theor
Comput Sci 2021;871:94–106. https://doi.org/10.1016/j.tcs.2021.04.014.
[23] Chaplick S, Kindermann P, Klawitter J, Rutter I, Wolff A. Morphing Rectangular Duals, 2023, p.
389–403. https://doi.org/10.1007/978-3-031-22203-0_28.

[24] Yeap K-H, Sarrafzadeh M. Floor-Planning by Graph Dualization: 2-Concave Rectilinear Modules.
SIAM J Comput 1993;22:500–26. https://doi.org/10.1137/0222035.
[25] Sun Y, Sarrafzadeh M. Floorplanning by graph dualization: L-shaped modules. Algorithmica
1993;10:429–56. https://doi.org/10.1007/BF01891831.
[26] Liao C-C, Lu H-I, Yen H-C. Compact floor-planning via orderly spanning trees. J Algorithms
2003;48:441–51. https://doi.org/10.1016/S0196-6774(03)00057-9.
[27] de Berg M, Mumford E, Speckmann B. On rectilinear duals for vertex-weighted plane graphs.
Discrete Math 2009;309:1794–812. https://doi.org/10.1016/j.disc.2007.12.087.
[28] Akbari Jokar MR, Shoja Sangchooli A. Constructing a block layout by face area. Int J Adv Manuf
Technol 2011;54:801–9. https://doi.org/10.1007/s00170-010-2960-4.
[29] Ueckerdt T. Geometric representations of graphs with low polygonal complexity 2012.
[30] Alam MJ, Biedl T, Felsner S, Kaufmann M, Kobourov SG, Ueckerdt T. Computing cartograms
with optimal complexity. Proc. twenty-eighth Annu. Symp. Comput. Geom., New York, NY, USA:
ACM; 2012, p. 21–30. https://doi.org/10.1145/2261250.2261254.
[31] Chang Y-J, Yen H-C. Rectilinear Duals Using Monotone Staircase Polygons, 2014, p. 86–100.
https://doi.org/10.1007/978-3-319-12691-3_8.
[32] Chang Y-J, Yen H-C. Constrained floorplans in 2D and 3D. Theor Comput Sci 2015;607:320–36.
https://doi.org/10.1016/j.tcs.2015.07.063.
[33] Pinki P, Shekhawat K. Linear-time construction of floor plans for plane triangulations. Commun
Comb Optim 2023;8:673–92. https://doi.org/10.22049/cco.2022.27814.1361.
[34] Pinki P, Shekhawat K. Characterization of Graphs Based on Number of Bends in Corresponding
Floor plans. Proc. 6th Int. Conf. Algorithms, Comput. Syst., New York, NY, USA: ACM; 2022, p.
1–8. https://doi.org/10.1145/3564982.3564994.
[35] Shekhawat K, Lohani R, Dasannacharya C, Bisht S, Rastogi S. Automated generation of floorplans
with non-rectangular rooms. Graph Models 2023;127:101175.
https://doi.org/10.1016/j.gmod.2023.101175.
[36] Baybars I, Eastman CM. Enumerating architectural arrangements by generating their underlying
graphs. Environ Plan B Plan Des 1980;7:289–310. https://doi.org/10.1068/b070289.
[37] Nummenmaa J. Constructing compact rectiliner planar layouts using canonical representation of
planar graphs. Theor Comput Sci 1992;99:213–30. https://doi.org/10.1016/0304-3975(92)90349-K.
[38] Wang X-Y, Zhang K. Generating layout designs from high-level specifications. Autom Constr
2020;119:103288. https://doi.org/10.1016/j.autcon.2020.103288.
[39] Bisht S, Shekhawat K, Upasani N, Jain RN, Tiwaskar RJ, Hebbar C. Transforming an Adjacency
Graph into Dimensioned Floorplan Layouts. Comput Graph Forum 2022;41:5–22.
https://doi.org/10.1111/cgf.14451.
[40] Raveena, Shekhawat K. A theory of L-shaped floor-plans. Theor Comput Sci 2023;942:57–92.
https://doi.org/10.1016/j.tcs.2022.11.025.
[41] Merrell P, Schkufza E, Koltun V. Computer-generated residential building layouts. ACM
SIGGRAPH Asia 2010 Pap. - SIGGRAPH ASIA ’10, New York, New York, USA: ACM Press;
2010, p. 1. https://doi.org/10.1145/1866158.1866203.
[42] Rodrigues E, Gaspar AR, Gomes Á. An evolutionary strategy enhanced with a local search
technique for the space allocation problem in architecture, Part 2: Validation and performance tests.
Comput Des 2013;45:898–910. https://doi.org/10.1016/j.cad.2013.01.003.
[43] Wu W, Fan L, Liu L, Wonka P. MIQP‐based Layout Design for Building Interiors. Comput Graph
Forum 2018;37:511–21. https://doi.org/10.1111/cgf.13380.
[44] Hu R, Huang Z, Tang Y, Van Kaick O, Zhang H, Huang H. Graph2Plan. ACM Trans Graph
2020;39. https://doi.org/10.1145/3386569.3392391.

[45] Laignel G, Pozin N, Geffrier X, Delevaux L, Brun F, Dolla B. Floor plan generation through a
mixed constraint programming-genetic optimization approach. Autom Constr 2021;123:103491.
[46] Fan Z, Liu J, Wang L, Cheng G, Liao M, Liu P, et al. Automated layout of modular high-rise
residential buildings based on genetic algorithm. Autom Constr 2023;152:104943.
[47] He F, Huang Y, Wang H. iPLAN: Interactive and procedural layout planning. Proc. IEEE/CVF
Conf. Comput. Vis. Pattern Recognit., 2022, p. 7793–802.
[48] Wu W, Fu X-M, Tang R, Wang Y, Qi Y-H, Liu L. Data-driven interior plan generation for
residential buildings. ACM Trans Graph 2019;38:1–12. https://doi.org/10.1145/3355089.3356556.
[49] Lu Z, Wang T, Guo J, Meng W, Xiao J, Zhang W, et al. Data-driven floor plan understanding in
rural residential buildings via deep recognition. Inf Sci (Ny) 2021;567:58–74.
https://doi.org/10.1016/j.ins.2021.03.032.
[50] Para W, Guerrero P, Kelly T, Guibas L, Wonka P. Generative Layout Modeling using Constraint
Graphs. 2021 IEEE/CVF Int. Conf. Comput. Vis., IEEE; 2021, p. 6670–80.
https://doi.org/10.1109/ICCV48922.2021.00662.
[51] Nauata N, Hosseini S, Chang K-H, Chu H, Cheng C-Y, Furukawa Y. House-GAN++: Generative
Adversarial Layout Refinement Network towards Intelligent Computational Agent for Professional
Architects. 2021 IEEE/CVF Conf. Comput. Vis. Pattern Recognit., IEEE; 2021, p. 13627–36.
https://doi.org/10.1109/CVPR46437.2021.01342.
[52] Klawitter J, Klesen F, Wolff A. Algorithms for Floor Planning with Proximity Requirements, 2022,
p. 151–71. https://doi.org/10.1007/978-981-19-1280-1_10.
[53] Duarte JP. Customizing mass housing: a discursive grammar for Siza’s Malagueira houses 2001.
[54] Müller P, Zeng G, Wonka P, Van Gool L. Image-based procedural modeling of facades. ACM
Trans Graph 2007;26:85. https://doi.org/10.1145/1276377.1276484.
[55] Wu F, Yan D-M, Dong W, Zhang X, Wonka P. Inverse procedural modeling of facade layouts.
ACM Trans Graph 2014;33:1–10. https://doi.org/10.1145/2601097.2601162.
[56] Rahbar M, Mahdavinejad M, Markazi AHD, Bemanian M. Architectural layout design through
deep learning and agent-based modeling: A hybrid approach. J Build Eng 2022;47:103822.
https://doi.org/10.1016/j.jobe.2021.103822.
[57] Wang L, Liu J, Zeng Y, Cheng G, Hu H, Hu J, et al. Automated building layout generation using
deep learning and graph algorithms. Autom Constr 2023;154:105036.
[58] Weber RE, Mueller C, Reinhart C. Automated floorplan generation in architectural design: A
review of methods and applications. Autom Constr 2022;140:104385.
[59] Watson KH, Giffin JW. The Vertex Splitting Algorithm for facilities layout. Int J Prod Res
1997;35:2477–92. https://doi.org/10.1080/002075497194615.
[60] Upasani N, Shekhawat K, Sachdeva G. Automated generation of dimensioned rectangular
floorplans. Autom Constr 2020;113:103149. https://doi.org/10.1016/j.autcon.2020.103149.
[61] Shekhawat K, Jain RN, Bisht S, Kondaveeti A, Goswami D. Graph-based approach for
enumerating floorplans based on users specifications. Artif Intell Eng Des Anal Manuf
2021;35:438–59. https://doi.org/10.1017/S0890060421000275.
[62] Wong DF, Liu CL. A New Algorithm for Floorplan Design. 23rd ACM/IEEE Des. Autom. Conf.,
IEEE; 1986, p. 101–7. https://doi.org/10.1109/DAC.1986.1586075.
[63] Seebohm T, Chan D. The design space of schematic Palladian plans for two villa topologies.
Reinventing discourse-how Digit. tools Help Bridg. Transform Res. Educ. Pract. Archit. Proc.
twenty first Annu. Conf. Assoc. Comput. Des. Archit., 2001, p. 156–65.

Generation of Dimensioned Floor Plans for a Given Boundary Layout

More Related Content

Similar to Generation of Dimensioned Floor Plans for a Given Boundary Layout (16)

More from Journal of Soft Computing in Civil Engineering (20)

Recently uploaded (20)

Generation of Dimensioned Floor Plans for a Given Boundary Layout