JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J.

IASS

NEXT GENERATION PARAMETRIC DESIGN

JEROEN L. COENDERS
CEO/CTO/Founder, White Lioness technologies, Oostenburgergracht 79, 1018 NC Amsterdam, The Netherlands,
[email protected]

Editor’s Note: Manuscript submitted 15 February 2021; revision received 19 May 2021; accepted 02 June 2021. This
paper is open for written discussion, which should be submitted to the IASS Secretariat no later than December 2021.

DOI: https://doi.org/10.20898/j.iass.2021.018

ABSTRACT
This paper presents a novel, next-generation, cloud-native parametric and associative platform for digital
knowledge, services and automation, and the rationalisation behind the development of and the need for this
platform in relation to the history of computational design and engineering, and the advantages and limitations
of each step in this evolution: Computer Aided Design (CAD), Building Information Modelling (BIM), Finite
Element Analysis (FEA), Parametric and Associative Design (PAD), Generative Design and programming
approaches to design and engineering. The paper discusses some of the key functionalities in relation to why
they are useful as a next step in the digital transformation of the Architecture, Engineering and Construction
(AEC) industry. The paper concludes with some of the challenges for the near future of this platform.

Keywords: Parametric design, Building Information Modelling, Computer Aided Design, Generative Design,
programming, software development, digital transformation

1. INTRODUCTION such as analysis and simulation, and three-

dimensional (3D) modelling in particular can greatly
At least three decades architects, (structural)
enhance the design and engineering qualities by
designers, engineers and contractors have been
being able to predict the outcome of construction, the
exploring to employ computational technology to
construction process (4D, 5D, and beyond), and the
improve the efficiency of the Architecture,
resulting performance of the object constructed.
Engineering and Construction (AEC) industry. This
paper presents a novel, next-generation, cloud-native However, when considering the mainstream
parametric and associative platform for digital software application typologies which have arisen
knowledge, services, and automation [1] and the over the years, such as Computer Aided Design
rationalisation behind the development of and the (CAD), Finite Element Analysis (FEA) and Building
need for this platform in relation to the history of Information Modelling (BIM) [2], have only
computational design and engineering, and the delivered limited results on this vision. Also
advantages and limitations of each step in this approaches which have had less adoption throughout
evolution. the industry such as programming approaches and
Generative Design [3] have proven to only deliver
The first assumption in employing digital technology
limited contributions to the promise of the
has been that before performing the physical acts of
computation vision of a more efficient and effective
construction, virtual design, analysis, and
industry.
construction would improve the quality of the
process and end result (the built object in its 2. COMPUTER-AIDED DESIGN (CAD)
lifecycle) as well as lower the cost due to less failure
cost and faster execution. Regular (non- Computer Aided Design [4] is one of the first
computational) design and engineering activities developments in virtual design and has enhanced and
such as designing, drafting and calculation are also improved the original manual drafting process which
forms of these virtual activities. However, the second technicians undertook and which involved drawing
assumption has been that computational technology, board, ruler, pen and paper by replacing it by

Copyright © 2021 by Jeroen L. Coenders.

Published by the International Association for Shell and Spatial Structures (IASS) with permission. 153
 Vol. 62 (2021) No. 2 June n. 208

originally two-dimensional computational drafting However, although the introduction of Building

(2D CAD) and later three-dimensional geometrical Information Modelling has delivered great
modelling applications (3D CAD). This information quality improvements over Computer
development has sped up the process of designing Aided Design, the systems unfortunately have not
buildings and other works and has simplified the been able to deliver the complete promise of the
process of adjustment of the design compared to the vision of virtual design and construction.
manual labour of drafting on paper. However, the
One of the key issues is that Building Information
technology was (originally) limited to geometrical
Modelling systems are able to describe the desired
representation of the design only, leaving out many
object to be constructed with rich information
crucial aspects of the information of the object to be
through their object definition structure (object
constructed undefined or to be defined in other
oriented) but that this object definition structure does
places leading to discontinuous information storage
not cater for analysis or simulation information, or
of the design information. The weakness in this
only very limited, or design and engineering
manual process therefore lies in the fact that the
(justification) information. In other words, Building
coordination and consistency of information is error-
Information Modelling is able to describe a rich
prone and labour intensive due to the distributed
product model of what needs to be built and what
nature of information in multiple places and reliance
properties this object needs to have but does not
on manual coordination activities to align the
provide any information about ‘the why’: for
information between the different information
example, why is the object dimensioned as it is? Or
carriers.
why is a particular material chosen? Similar to the
disadvantages of the Computer Aided Design
3. BUILDING INFORMATION MODELLING systems processes the types of information which
(BIM) cannot be stored in the Building Information
Modelling system will have to be stored in other
Building Information Modelling [2] has further
disconnected locations and be kept aligned often
improved the virtual design process of buildings and
manually which is error-prone and labour intensive.
other objects for construction by adding object
definitions to the information model which describes Another issue is the difference in nature between the
the object to be constructed. Examples of Building information which is produced in design and
Information Modelling systems are for example engineering processes and the information which is
Revit [5], Tekla [6] and ArchiCAD [7]. required for construction. Building Information
Modelling systems initially were built for
In Building Information Modelling the three-
construction. The information required for
dimensional representation of the object is no longer
construction which is stored in Building Information
just geometrical (as is the case for CAD), but three-
Modelling systems expresses the desire of what the
dimensional geometry becomes a derivative of an
client (or engineer/contractor) would like to be built
object definition which is able to also describe
and should therefore be unambiguous and preferably
(other) properties (characteristics) of the object as
highly detailed as no misunderstanding can take
well as relationships to other modelled objects.
place then. However, in the case of design and
Furthermore, (limited) behavioural characteristics
engineering processes often many options,
can be part of the object definition. The object’s
alternatives, cases, combinations, etc. are explored
properties can represent all kinds of useful
by the designer or engineer which together provide a
information of the object modelled such as its type,
design justification behind the choices which the
material, material properties and relationships to
designer or engineer has made about the design.
other objects such as connectivity and location.
Furthermore, many different cases, scenarios and
Because these objects are contained in a database
situations are analysed by designers and engineers to
querying the model for information becomes
prove the soundness of the design.
possible which for example allows for automated
coordination between the objects through clash- A paradox exists between this type of design
detection, automated quantity take-off and rule- information and the information for construction
checking. With these cost estimation and which leads to differences in the requirements for
construction scheduling analysis becomes within systems which would need to store this information.
reach too. Building Information Modelling only caters for the

154
 JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J. IASS

latter: detailed unambiguous object-oriented to be gathered from the physical reality as the virtual
information. But design and engineering requires information on the object is lost over time.
abstract information and multiple statements about
the same subject to be stored. 4. FINITE ELEMENT ANALYSIS (FEA)

The second paradox which emerges is the growing Alongside the development and introduction of CAD
nature of information during the design and and BIM systems another type of systems emerged
engineering process. At the early stages of the which has helped designers and engineers to
development of an object often very little virtually assess the predicted performance of the
information exists. Over the course of the design and objects constructed: analysis and simulation tools.
engineering process (these days augmented by
computational tools) the information grows in One of the most known categories of tools is Finite
volume and complexity as options and alternatives Element Analysis (FEA) which can (amongst others)
are explored, choices are being made and more be used by structural engineers to assess the
detailed information on the (predicted) performance structural performance of an object in terms of stress,
of the designed object becomes available. The object deformations, dynamics, etc.
definition of Building Information Modelling Many examples of these systems exist as academic
however often requires detailed choices about what or commercial applications, such as Sofistik, GSA,
object to use and what its properties are. At the early Ansys, Diana, SCIA, RFEM, MatrixFrame, Strand,
stages this information is unknown as it is the Robot, and Etabs.
outcome of the above-described design process, and
therefore still to be determined and not available yet. Although these tools have been widely adopted and
The information which leads to these choices can are very useful to produce analysis and simulation as
often not be stored in Building Information a basis for explanation and justification about the
Modelling systems and the early choices are lost as design and engineering rationale behind the object to
they often are abstract or do not possess the required be constructed, unfortunately their disconnection
level of detail (that of an object). Again, as this from other tools such as CAD and BIM systems
information is lost or has to be kept in another unfortunately has not helped to overcome the earlier
location, the risk is misalignment and laborious issues around misalignment and laborious
manual coordination processes emerge. coordination.

Another aspect which has to be mentioned is that It has to be mentioned that while CAD and BIM
design and engineering information is not only systems often do not contain much functionality for
valuable for the construction process to prevent information on alternatives or options to be stored (as
failure cost, but also in the future lifecycle in the mentioned because they were not built for this
building, for example for facility management and purpose), especially Finite Element Analysis tools
changes in the design at a later stage. Buildings often often have functionality to study different scenarios
exist for decades and often have multiple uses during such as load cases, load combinations and varying
their lifecycle. Conversions, refurbishment or support conditions bringing them closer to the design
renovation often are required to keep the building up and engineering justification process the designer or
to date with the latest requirements from clients and engineer requires for the design and engineering
tenants. Design and engineering information often process of the object constructed. Unfortunately, the
contains crucial information on why certain options and alternatives often do not involve
components of the building exist and why they have geometry or other physical properties than the ones
the properties that they have. This is very valuable mentioned above.
information when the building or other work needs At times many analysis and simulation tools have
to be changed. Building Information Models deliver loose interoperability (the ability to exchange data
on the promise that they can provide the information and/or information) with CAD or BIM systems, but
on what is there, but unfortunately not why. This another issue of potential misalignment emerges
information needs to come from other disconnected here and causes misalignment risks: that of the use
places and often cannot be retrieved after many of schematisation. Analysis and simulations often
years. This often leads to rework in design and require some kind of simplified schematisation
engineering, expensive investigations and at times compared to the (detailed geometrical) information
unnecessary demolition works as the information has stored in CAD or BIM systems which contain the full

155
 Vol. 62 (2021) No. 2 June n. 208

geometry and related information. These 5. DESCRIBING DESIGN LOGIC THROUGH

schematisations, such as idealised centerline SPREADSHEETS OR PROGRAMMING
geometry, meshes or prismatic sectional geometry,
One of the challenges to overcome the above-
somehow have to be derived from and/or translated
mentioned misalignment would be to describe
into the full geometry of the object to be constructed
design and engineering logic as part of the integral
described in the CAD or BIM system. For
information model of the object to be constructed. As
computers this is not a trivial task. For a human
demonstrated CAD (computational geometry) and
design or engineer such a translation is quite easy
BIM (three dimensional object-oriented description)
to make and judge, and are part of the basic training
have not been able to form this integral information
of every engineer. However, computer applications
model due to their restrictions to describe the logic
struggle with (automatically) interpreting these
(and justification) of the design.
translations and reasoning about the correct
justification of the schematisation. The rigid nature For decades designers and engineers have employed
of the translations in these applications often do not other means to describe this design and engineering
lead to the schematisation a human engineer would logic by making their own tools, such as by using
use in his or her design justification as this often spreadsheets and using computer programming
involves reasoning about conservative (safe) languages.
approaches and are dependent on the global
problem (the total system) instead of just the local Spreadsheets, such as Microsoft Excel [8], allow the
translation of the object at hand. designer or engineer to define data and express
formulas to work with this data to automate some of
For example, when modelling a continuous beam her work. This has allowed the designer or engineer
for a reinforcement dimensioning calculation the to encode calculations and analysis procedures to
schematisation of the supports influence the speed up some of her activities. This can be seen as
outcome of the bending moment and shear force a good step towards expression of design and
distributions along the beam which are used to engineering logic in a computational information
estimate the reinforcement quantity required for model.
structural safety. In case an engineer would like to However, spreadsheets are rather limited in the type
make a conservative estimate for both the of logic they allow the engineer to express. The
reinforcement over the supports as well as the functionalities are usually limited to numerical and
reinforcement in the field no single support mathematical operations and simple data handling
condition schematisation of the beam gives a functionalities. Furthermore, the table structure
conservative approach for both situations. (rows and columns) of spreadsheets limits the type
Therefore, multiple support conditions should be of data structure spreadsheets can handle to simple
analysed to come to the correct conservative numbers, lists or matrices. More complex data
estimate of the reinforcement in both locations. So structures are difficult to handle. Next to the
at least two models need to be analysed or a model limitations in the structure and volume of the data
with two support condition cases. which spreadsheets allow designers and engineers to
handle also this approach suffers from loose
For the systems involved this means that not only interoperability with other information model
translation of one geometry into another carriers. The information in spreadsheets and other
schematised geometry (and related properties) is information systems, such as CAD, BIM and/or
involved but that this translation is dependent on the FEM needs to be manually coordinated and aligned
desired outcome in the justification process that is as mentioned error-prone and laborious.
part of the design and engineering process. So far,
to the best knowledge of the author, none of the Another means which some designers and engineers
tools involved have been able to store this (the author being amongst them) have employed from
information in its entirety and approaches which a very early stage, is to use computer programming
use loose interoperability again suffer from the risk languages to encode design and engineering logic in
of misalignment and laborious and error-prone homegrown computer applications which either
coordination processes to mitigate these effects on generate the object or perform automation tasks
changes in the design of the object to be similar to what designers and engineers use
constructed. spreadsheets for. Computer programming however

156
 JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J. IASS

lifts some of the limitations which spreadsheets suffer looking at the field of software development, it can
from. Computer programming languages often have be observed that this field does not consist of a single
far richer means to express logic than spreadsheets as role (‘the computer programmer’ as it perhaps was
well as provide toolkits in the form of frameworks and in the early days of computers), but many specialistic
libraries with many advanced functionalities. roles: front-end developers, back-end developers,
full-stack developers, database developers, cloud-
Secondly, computer programming languages allow
native developer, DevOps developers, testers,
for richer data structures, such as lists, data trees,
quality assurance, (database or infrastructure)
dictionaries, hash tables, etc. and can store and query
operators, data scientists, artificial intelligence
this data in more advanced file structures or
specialists, security specialists, etc. Many of these
databases. This allows the programmer to express
roles do not even emerge from a single education and
logic in a far more powerful manner than
keeping up with all developments in these fields is
spreadsheets allow.
an enormous if not impossible task. Computer
Thirdly, the resulting computer applications often science is an academic field on its own with many
are able to handle larger data sets more easily and subfields.
more efficiently than spreadsheets.
So, as a designer or engineer to become a specialist
Fourthly, the resulting computer applications often in the fields related to designing and engineering
do not have to sit on their own but can connect to AEC objects (also many fields on their own) as well
other applications through API’s (Application as to become a specialist in software development is
Programming Interfaces) which allow them to virtually impossible (perhaps except for the
retrieve data from other applications or send data to exceptionally gifted). In the experience of the author
other applications and use the functionality in other the combination of two specialisms in a single person
applications. often leads to mediocreness in two fields rather than
Finally, the programmer has very little limitations to becoming a specialist. Of course, perhaps a new role
what can be built by making use of the programming could emerge where the designer or engineer
languages. All aforementioned applications such as complements his knowledge and skills related to
CAD and BIM are a result of computer programming designing and engineering the AEC objects with a
too. Any designer or engineers could, in theory at limited set of skills in programming as a tool, but the
least, invent their own application which is fit for question then becomes if some of the
purpose to the needs of the project or task. aforementioned advantages of computer
programming then still apply. Especially when
While computer programming seems the most developing more complex reusable systems it is
powerful approach and the approach that delivers the rather dangerous when the designer or engineer does
most freedom to the designer or engineer to achieve not understand well how to properly build such a
the aforementioned computational vision, one might system and how to ensure its safety and validity. This
wonder why designers and engineers should not all could potentially lead to grave mistakes and
learn computer programming as a new skill. potentially accidents.
Unfortunately, this approach also has some severe
drawbacks as the author has found out over his years Secondly, an approach could be chosen where
in a large global engineering firm. Many of his peers designers and engineers collaborate with computer
in other firms have similar experiences when programmers to develop the applications or solutions
experimenting with forming groups of designers and in multi-disciplinary teams. This is a realistic
engineers with programming skills. possibility and can be observed to function in some
companies. However, the question is if this will scale
Firstly, most designers and engineers are not to the entire industry. Computer programmers are
schooled as programmers or computer scientists. scarce, difficult to retain and expensive for AEC
They often have an education with either an firms who have to compete with data science,
architectural, building technology, structural artificial intelligence and cloud computing
engineering or civil engineering background of companies for talent and skills. Additionally, even in
which the curriculum often contains some basic tight-knit collaborative teams working with multiple
programming, but not a full education in computer roles requires extensive communication which often
science and the software development process proves to be a source of error and misunderstanding.
(which are already two fields on their own). When The speed involved often is rather low to other

157
 Vol. 62 (2021) No. 2 June n. 208

approaches where the designer or engineer is more development suffer from the above-mentioned
in control and has a deeper understanding. Often limitations and risks.
designers and engineers do not trust the tools built
6. PARAMETRIC DESIGN
and this leads to low adoption and use in practice,
which is rather wasteful for the companies who made About 18 years ago a new generation of parametric
the investment. and associative design systems emerged [9] which
has provided a (partial) solution for the further
Thirdly, programming uses codification to encode
expression of design and engineering logic and
the design and engineering logic into source code.
justification. In order of appearance
The word ‘code’ says it all: it requires you to learn
GenerativeComponents [10], Grasshopper [11] and
and understand the code language which is not the
Dynamo [12] have emerged as parametric systems
native language you express yourself in. This throws
which share a few crucial characteristics: an explicit,
up an adoption barrier for many people which
visual logic graph with user-defined parameters and
sometimes is too difficult to overcome. The ‘code
user-defined relationships between the operations, a
language’ leads to an in-crowd who speaks the
single-directional graph solver which solves the
language but excludes others.
graph on change, the mechanism of replication,
Fourthly, the complexity of writing a simple extensibility (through packages, extensions or
automation routine to generate a piece of geometry plugins) [13] and community-driven support and
is not comparable to the complexity programmers development. Opposed to the earlier textual
encounter when writing automation between programming languages these systems use visual
applications or when scaling over the size of a single programming languages in order for the designer or
machine (such as is often required for distributed or engineer to express their design and engineering
parallel computing of real world multi-disciplinary logic. As the visual nature of these programming
optimization problems). The lack of knowledge on languages as well as the interactivity and feedback
software development of such complex automation mechanisms tightly couple to modelling geometry in
and interoperability cases often causes the resulting CAD and BIM systems and other outcomes of the
software to suffer from stability, robustness issues or logic, this type of programming languages is far
even simply providing the wrong outcomes. easier for designers and engineers to adopt than their
textual siblings. If studied the nature of the
Fifthly, coding is a rather slow process when done
interaction of these systems and the workflow of the
properly. Small applications often have little
user is very similar to how spreadsheets are being
development overhead for setting up development
used by designers and engineers to build their own
and testing environments (but often are not
automation logic: highly interactive and responsive
reusable), but when applications grow larger, more
to the steps the designer or engineer takes.
complex or developed with multiple people in teams
(as is often the case in real world problems or when The expression of logic in these systems can be
trying to scale solutions to the entire organisation), thought of as composition of digital building blocks
in order to maintain the solution and control a proper where each of the blocks express reusable operations
outcome, more overhead is required to be invested to which can be composed in various arrangements to
set up these environments. Very often projects will build algorithms.
encounter that the software produced for the project
Initially, parametric design was mainly used for
is ready too late to be used in the project and project-
modelling (controlling) geometry due to the close
overarching carry-over and maintenance which
relationships between these systems and advanced
makes the tools useful on multiple projects is too
3D CAD systems which support advanced types of
expensive.
geometry, such as Rhinoceros3D [14]. However,
It needs to be noted that the author does not oppose nothing in these systems limits their use to geometry-
the use of programming for any type of software only. The systems contain advanced data and object
development. The use of programming for types on which operations can take place as well as
applications which are to be reused by large numbers advanced data structures, such as data trees.
of people, throughout an organisation or through the Furthermore, the mechanism of replication allows
industry even, often requires programming users to express logic for a single entity once but
approaching, but for the expression of design and allow it to be replicated over a multitude of entities
engineering logic programming and software when applied to a data structure containing multiple

158
 JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J. IASS

entities. The systems link closely to BIM systems Although the many advantages of the current
[15] (often through plugins) or in some cases even applications of parametric design unfortunately the
live within these systems, such as is the case for systems still also struggle to deliver the full promise
Dynamo in Revit [12] and Rhino Inside Revit [16], of the computation vision. This is mainly caused by
and these systems are able to work directly with and the limitations which the computational power of a
on the objects expressed in BIM systems. single machine imposes on the automation. It turns
out that in the real-world full models of buildings and
Compared to BIM systems, the design and other objects to be constructed with all their design
engineering logic in the case of Parametric Design and engineering logic, alternatives, options,
can overarch the single object on a much wider scale connected simulations and analyses become rather
and can even be expressed and controlled large and therefore difficult to process for a single
independently from physical or semi-physical machine. Also, in case of interconnected software
objects (semi-physical objects in BIM are for applications through the mentioned extensions or
example ‘rooms’) as BIM requires. This means that plugins or API’s, the automation suffers from
the logic can be both dependent and independent instability. Often this has very little to do with the
from the detailed objects and therefore is much more applications themselves, but the underlying
usable for expression of (more abstract) early-stage operating system and the means of communication
design and engineering logic. provided. These communication means seem not to
be built for such intensive integrations and the
Furthermore, parametric design allows the designer stability required by the automation.
or engineer to express her design or engineering
activities in a way which is very close to the often Another limitation is the parametric software itself.
algorithmic nature of these activities. Building codes Especially when sharing built applications with
and calculation procedures often follow a protocol of others, the other person needs to know how to
steps and checks to prove that the object constructed operate the parametric software and the model in
is safe. order to use it. This at times proves to be challenging
in practice.
Also, parametric design allows the designer or
engineer to express multiple situations as are The final limitation which needs to be mentioned is
involved in a design or engineering justification of reporting. Although parametric design is able to
the object to be constructed. contain much of the design and engineering logic,
communicating this information to the rest of the
Tight integrations exist with finite element software world in the form of structured reports (as is common
both by connecting between the parametric software practice in the industry) does not seem one of its
and the finite element software, or by integration of features. Information therefore has to be carried to
the finite element method in the parametric software other applications (suffering from the above-
itself as an extension [17]. mentioned interoperability and automation issues) in
Next to the aforementioned advantages, order to achieve this.
development in parametric design is relatively fast Concluding it can be said that parametric design is a
compared to programming approaches and is far huge step forward in the realisation of the
more insightful to the designer or engineer in computational vision, but also still suffers from quite
absence of a codified language. Special skills besides a few limitations.
basic modelling skills, some understanding of logical
steps and a small introduction in the building blocks 7. OPTIMISATION AND GENERATIVE
is not required. The interaction the systems provide DESIGN
between the act of modelling a piece of logic and the
direct feedback provides a lot of insight and lowers Optimisation by making use of computers is a field
the complexity of translation of the logic in which precedes parametric design but has become
parameters, operations and the proper relationships. more accessible through parametric design
This means that (expensive) programmers are not applications. In optimisation an objective function is
required and designers or engineers can easily pick either minimised or maximised under a set of
this type of expression of logic up beside their constraints in order to automatically achieve the
normal work and specialty. desired performance of an object.

159
 Vol. 62 (2021) No. 2 June n. 208

Before parametric design applications emerged the Element method [24] is employed encoded in pre-
designer or engineer had to rely on her own software programmed parametric building blocks. This
development in order to employ automated method is fast but less precise than FEM, but for
optimisation techniques in her design process. early stages design purposes fit for use. The vision
Various applications [18] and frameworks [19] have for these building blocks is to form a library of
been developed to overcome these limitations, but (structural) components which the designing
they never have become widely adopted in the AEC engineer can easily use in a parametric system to
practice. quite come to a rough performance assessment of the
building at the earliest possible stages in the project:
With the emergence of Grasshopper’s Galapagos
preferably at the first meeting with the design team.
[20] and other plugins, optimisation has become far
The application of a library of parametric
more accessible as optimisation (Genetic Algorithms
components is not necessarily new (all parametric
and Simulated Annealing in the case of Galapagos)
systems have some sort of library of building
now can be used as a building block as envisioned in
blocks), but StructuralComponents’ concepts around
the author’s openStrategy Form Finding framework
employing large building blocks with all design logic
[19]. Galapagos has delivered on this vision and
embedded and making these available in a library for
other plugins have emerged which implement other
use in the earliest stages is an important development
algorithms.
worth mentioning. However, it needs to be
The combination of parametric design and mentioned that while some experiments in practice
optimisation techniques, mainly topology have taken place, adoption in practice is rather low
optimisation [21], is now popularly referred to as at the moment since the software is not generally
Generative Design [3]. available.
Optimisation only delivers a small part of the 9. NEXT GENERATION SYSTEM:
computational vision and does not deliver any of the PACKHUNT.IO
capabilities which the other mentioned approaches
All aforementioned developments in the
have brought. However, it is a useful and advanced
computational vision have delivered a lot of
functionality which will bring the industry much as
functionality but still impose limitations in the area
themes like sustainability and performance become
of availability, accessibility and scalability (with
more important. Optimisation can potentially aid to
automation robustness, stability and interoperability
minimise material and energy use while maximising
as important themes). To overcome these limitations
the performance of the objects constructed.
the author and his team have over the last 7 years
Although parametric design systems have made it built Packhunt.io [1], which is a cloud-native
easier to apply optimisation techniques for the platform for parametric and associative design,
engineer without (textual) programming skills, the automation and digital services.
application of optimisation at the moment is severely
Cloud-native means that the system uses
limited as optimisation requires much computational
components from the cloud (many computers in data
power (the search spaces involved in real world
centers) but also its abilities to scale with the amount
problems are extremely large). Distributed
optimisation is experimented with [22] but at the of work that the system will have to do. Scaling
means that more computers are employed as more
moment is not available to the general public and
work needs to be done. This amongst others follows
suffers from interoperability issues as no common
principles of distributed and parallel computing but
standard for communication between applications
also the use of autoscaling (scaling automatically)
exists.
and the use of a microservices architecture.
8. STRUCTURALCOMPONENTS
Parametric and associative languages, such as the
StructuralComponents [23] is a research area where Grasshopper3D visual language, are used as building
multiple students of Delft University of Technology blocks to express automation logic which is referred
have built on each other’s work to develop an to as ‘digital knowledge’ or ‘digital services’. Also,
alternative approach to early stage analysis textual programming languages can be used to
capabilities in parametric design systems. Instead of express automation logic either inside the parametric
employing FEM to perform analysis the Super models or outside these models in the form of

160
 JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J. IASS

microservices. Additionally, to employing building or have its own APIs called. This means that
blocks to express design and engineering logic in functionality existing in other systems does not have
parametric models, building blocks are used to to be duplicated but can be reused.
express all functionality in the solutions built with
As the platform is cloud-based, web-based and
this platform. For example, the front-end (web-based
online (accessed through a (modern) web browser
user interface) layout and the interaction on the front-
such as Chrome) this means that there is no
end which the user interacts with are expressed as
dependence on other software applications or
building blocks to build rich user interfaces which
devices and therefore the barriers to adoption by a
can follow workflows, present models and
wide audience are low.
information, etc. Also back-end processes to
automate complex processing of data and databases The platform allows for automation tasks, single or
are available through building blocks. This principle in orchestrated chains of chains, but making use of
of using visual programming and building blocks is parametric solvers, microservices or connected
often referred to as ‘No Code’ (although it needs to systems which live within or outside the cloud
be mentioned that ‘Low Code’ and ‘Code’ infrastructure. Orchestration allows for a stable
approaches in C#, VB.Net, Python and Javascript are automation where potential failures of individual
also allowed). steps in the process can be automatically corrected to
The platform is a complex hybrid application which ensure a stable (fully completed) outcome.
consists of front-end and back-end technology which In case of analysis and simulation automation tasks,
interact through various API’s, such as websockets the platform integrates with e.g., FEA systems to
and RESTful APIs in which each part has its own perform the required analysis and allows the results
role: the front-end technology provides a rich to be automatically reported in e.g., templated PDF
interactive environment for high configurable user documents.
interfaces while the back-end technology takes can
of all the heavy lifting in terms of automation, Packhunt.io allows users to expose their solutions to
generation, solving and data processing. other users as ‘digital services’ similarly to a
building block in Grasshopper but in this case the
Multiple solvers are employed which can solve the operation runs remote on the infrastructure and only
parametric logic. Amongst others, Packhunt.io was exposes an interface (input and output) which the
the first platform in the world to support the supplying user chooses to use. These services can be
Rhinoceros3D 7 Rhino.Compute service [25] offered free of charge but also can be ‘sold’ to other
(Rhino7 solver) which is used as a solver for native users by making micropayments in order to retrieve
Grasshopper models but also contains a solver which some of the investment made in building, exposing
reads the Grasshopper language based on the open and maintaining the service by the supplier of the
source Open Cascade geometry kernel [26] and knowledge.
which has more scalable characteristics and is more
performant. On the Rhino7 solver plugins can be By chaining different services together in a solution,
used as well as scripted components. digital supply chains can be built where the design
and engineering knowledge of an object to be
An example of such a plugin is Geometry Gym IFC constructed can be supplied in a distributed manner
[27] which allows the user to define logic to work rather than by a single effort. This is important as
with the OpenBIM IFC standard. Also, other knowledge in the AEC industry is complex and
connections to BIM applications, such as Revit and spread out over the industry. Nobody owns all the
Tekla are available for use. information and knowledge to build a building or
Building blocks are available in many different works with all its parts to the full detail.
areas, such as databases and files (storage of
Scalability is an important concept part of the
structured and unstructured data), emailing,
platform which is novel compared to previous
rendering, orchestration of data processing pipelines,
systems: the ability to employ more computers when
geospatial data processing, maps, charts, diagrams,
the work or the data volume or the parametric model
reporting, etc.
size surpasses the abilities of a single computer.
The platform is able to connect to other systems There are several reasons for this concept to be
through being able to call the API of other systems included.

161
 Vol. 62 (2021) No. 2 June n. 208

Firstly, the solutions built often involve multiple file-to-factory: direct generation of the production
users, sometimes on a public website. If the solution files which factories can use to produce the different
would rely on a single model and/or a single solver building components.
either the solution would not be responsive to
multiple users at the same time or the users would
have to be placed in a queue and would have to wait
for their turn. This is not desired on the current
internet where users have little patience.
Secondly, in novel design environments, real world
buildings (or other works), their models and all the
knowledge and logic related to their designs are large
and complex, especially if full detail is required. The
models cannot only be automatically generated on
the platform, but also can undergo a fully automated
assessment (through analysis and simulation
software) and checking. The knowledge chains in
this case become long and complex at some point in
the design process which means that replaying the
design and engineering logic becomes a large task
for a single computer leading to undesirable
slowdowns in the design and engineering process.
By employing multiple computers at the same time
this process is sped up.
The platform is employed for generating entire
buildings especially for modular, industrial,
prefabricated and conceptual building designs and
their related engineering, production
(manufacturing) and construction processes. This Figure 1: Example of a parametric building which a
generation to the full detail requires scalable models. high level of detailed information
Thirdly, in some use cases large and complex data
sets need to be transformed in order to be useful in 10. DIGITAL KNOWLEDGE SERVICES
the design and engineering process, such as is for
Through its digital services model Packhunt.io
instance the case with geospatial data processing.
provides people with (digital) knowledge and people
The platform can be seen as a next generation with a need for (digital) knowledge with an app store.
parametric system, but also a next generation BIM Suppliers of knowledge encode their knowledge in
system. If the design logic is structured in objects parametric models, data sources or software
which expose properties, geometrical representation applications and can get revenues from the use of
and behaviour, compatibility with object orientation these services by others. Users of knowledge can
exists which allows exchange with existing Building discover these services and use them without having
Information Modelling systems. By employing the knowledge themselves or having to invest in
parametric and associative modelling the design and building, encoding and maintaining the knowledge.
engineering logic of objects can completely be Even the opportunity exists for discovering the
described and composed into a larger object to be answer to a problem before the user knows the
constructed, such as a building. This object can be problem exists. These aspects benefit both sides of
virtually tested and checked before physical the ecosystem: each side gets concrete value in
construction. return.
Part of the process of transforming a virtual design This is an important aspect to lower the barriers of
into a physical object is manufacturing. In the past adoption of digital technology in the AEC industry:
manual work had to be done in order to manufacture lowering the investments which need to be made in
the desired object. However, the platform allows the digitisation of the industry and increasing the

162
 JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J. IASS

affordability of technology. In theory the investment detail of detail, complexity and quantity.
will only need to be made once in the industry and
In this fully parametric model the components are
the knowledge can be reused many times. This
not directly modelled as geometry, but derive their
potentially can speed up the adoption of digital
inputs from other parts of the model which represent
technology and the computational vision.
other disciplines.
Companies can use digital services to build digital
To give an example, the dimensions of the core are
knowledge assets for their company where
based on the height of the building and the various
knowledge becomes reusable throughout the
functions the core houses: e.g. elevators and vertical
company but also can become an alternative business
transport of building services. The desired number of
model.
elevators derived from the functional plan of the
11. HIGHRISE BUILDING CASE building as well as the desired net floor area for each
of the functions in combination with occupancy rates
In Figure 1, Figure 2 and Figure 3 an example of
drives the layout of the core which in its turn drives
building is presented which represents the level of
the core dimensions.
detail which is required for a real building case for a
highrise project.

Figure 3: Detailed view on the building from Figure 1

and Figure 2 which shows the level of detail generated
for a realistic building case
From the example it can be easily reasoned that such
a parametric model which describes a full building
in full detail quickly leads to an enormous amount of
interconnected rules and data volume (because of the
number of objects).
In case not one building scenario but many scenarios
would need to be explored, such as in the case of
Multi-Disciplinary Optimisation this problem
multiplies in order of magnitude.
The platform overcomes this complexity and data
volume by employing the aforementioned
Figure 2: Example of the structure of a building
distributed computing approach to maintain realistic
generated through parametric design where the generation times.
structural components have been part of inputs of other Figure 4 depicts a schematic representation of how
disciplines, e.g. the number of lifts in the core, the building models are generated consisting of smaller
required column free spans from the net floor area
models (digital services).
13. POTENTIAL CONSEQUENCES AND
Figure 1 and Figure 2 give a representation of how
FUTURE WORK FOR THE AEC
many elements are involved in the facade and
structure of a realistic highrise project. Architectural A potential consequence of these developments is
elements and especially MEP components have not that the roles of people and companies will change
been shown in these images but have an ever higher because of the emergence of the technology. The

163
 Vol. 62 (2021) No. 2 June n. 208

author would like to propose that to build and preferably these tools should not be replaced but
maintain the knowledge and build new knowledge connected to but due to unavailability of open
on top of existing knowledge new roles will emerge. standards, open file formats, API’s or issues around
licenses integration sometimes is challenging.
Rather than wasting their lives on performing
Especially building physics tools, tools for small
repetitive tasks which machines are able to do,
scale problems and supplier-specific tools (such as
knowledge and invention are tasks for humans. For
anchor calculations) pose a real problem. The author
companies, especially in the Western world where an
would like to propose that adding API’s to these
aging population is a real problem, such technology
systems will solve the problem.
brings a way to build digital assets independent of
people who might leave the company. The company 4. CONCLUSIONS
will be able to survive based on its digital assets
This paper has presented a novel cloud-native
while the people are still required to expand the
parametric and associative platform for digital
assets. Furthermore, as stated technology can bring
services and automation. The paper has discussed the
new business models and become a differentiator:
step-by-step development of a computational vision
the company with the smartest services will be
which is shared by many in the Architecture,
picked over the competition. To cater for these
Engineering and Construction industry with for each
changes in society, the author would like to propose
step the progress made and the limitations
that education is redesigned around the skills
remaining. Furthermore, the possibilities of a next
required to build, encode and maintain digital
generation platform have been discussed as well as
knowledge: the designers and engineers need to
some of the consequences this platform has for the
know the fundamentals of the knowledge involved,
industry as well as some of the challenges and future
reasoning in terms of systems rather than projects,
work which remains.
and being able to control and validate processes how
to build and maintain (safe) systems. ACKNOWLEDGEMENTS
Next to the challenge around the availability and The author would like to acknowledge and thank all
skills of people, especially interoperability with who have had the same computational vision as the
existing software tools proves to be a real challenge author and who have contributed to realisation of this
to further advance this approach as many different vision in the different steps as well as the ones who
systems, often local to region or market exist, which contributed to the emergence of the Packhunt.io
are difficult to integrate. In the vision of the author platform over the last years.

Figure 4: Schematic representation of the building model on Packhunt.io which consists of multiple parametric models
(digital services)

164
 JOURNAL OF THE INTERNATIONAL ASSOCIATION FOR SHELL AND SPATIAL STRUCTURES: J. IASS

CALL FOR COLLABORATION [9] R. Aish, CustomObjects: a model-oriented

end-user programming environment. Draft
The author would like to call anybody who is paper submitted to the Visual End User
involved in parametric design and related workshop VL2000, 2000.
computational fields to join the IASS WG 13 Study
group on Next Generation Parametric Design and [10] R. Aish, Introduction to GenerativeComponents,
collaborate in this field. a parametric and associative design system
for architecture, building engineering and
Furthermore, the author would like to call anybody
digital fabrication. 2005, white paper,
who would like to expose his or her knowledge and
http://www.bentley.com, accessed 20 May
design and engineering logic to the rest of the world
2007.
so that we can lower the barriers to create a better
world with this knowledge to contact him to discuss [11] S. Davidson, Grasshopper3D website,
how to collaborate. https://www.grasshopper3d.com/, accessed
DISCLAIMER 14 February 2021.

This paper has been written by one of the founders [12] Autodesk, Dynamobim website,
and director of White Lioness technologies, which is https://dynamobim.org/, accessed 14 February
the company which develops the Packhunt.io 2021.
platform.
[13] J.L. Coenders, NetworkedDesign, PhD thesis,
REFERENCES Delft University of Technology, 2011.

[1] White Lioness technologies, Packhunt.io [14] Robert McNeel and Associates, Rhino3D
website, http://www.packhunt.io, accessed 14 website, https://www.rhino3d.com/, accessed
February 2021. 14 February 2021.
[2] C. Eastman, Building Product Models: [15] Trimble, Grasshopper-Tekla Live Link
Computer Environments Supporting Design website,
and Construction. CRC Press, 1999. https://teklastructures.support.tekla.com/not-
version-specific/en/ext_grasshopperteklalink,
[3] K. Shea, R. Aish, M, Gourtovaia, Towards accessed 14 February 2021.
integrated performance-driven generative
design tools, 2005, Automation in [16] Robert McNeel and Associates,
Construction. 14 (2): 253–264. (DOI: Rhino.Inside.Revit Beta website
10.1016/j.autcon.2004.07.002) https://www.rhino3d.com/inside/revit/beta/,
accessed 14 February 2021.
[4] D.T., Ross, Computer-Aided Design: A
Statement of Objectives, 1961. (DOI: [17] Karamba3D, Karamba3D website,
10.1145/366532.366554) https://www.karamba3d.com/, accessed 14
February 2021.
[5] Autodesk, Autodesk Revit website,
https://www.autodesk.com/products/revit/ [18] Phoenix Integration, ModelCenter website,
overview, accessed 14 February 2021. https://www.phoenix-int.com/, accessed 14
[6] Trimble, Tekla website, https://www.tekla.com/, February 2021.
accessed 14 February 2021. [19] J. Coenders, L. Wagemans, openStrategy
[7] Graphisoft, Archicad website, Form Finding, a new approach to structural
https://graphisoft.com/solutions/products/ form finding and structural optimization, in R.
archicad, accessed 14 February 2021. Metro, editor, Proceedings of the IASS2004
International Symposium Shell and Spatial
[8] Microsoft, Microsoft Excel website, Structures from Models to Realization,
https://www.microsoft.com/en-us/microsoft- Montpellier, France, September 20-24, 2004:
365/excel, accessed 14 February 2021. 66-67.

165
 Vol. 62 (2021) No. 2 June n. 208

[20] D. Rutten, Evolutionary Principles applied to Construction, Europia Productions, Paris

Problem Solving, blog post France, 2009: 491-500.
https://www.grasshopper3d.com/profiles/blo
gs/evolutionary-principles, 25 September [24] R. Steenbergen, Super elements in high-rise
2010, accessed 14 February 2021. buildings under stochastic wind load, PhD
thesis, Delft University of Technology,
[21] Burns, Recent advances in optimal structural 2007.
design, American Society of Civil Engineers,
2000. [25] Robert McNeel and Associates, Rhino
Developer Docs website,
[22] P. von Buelow, ParaGen: Performative https://developer.rhino3d.com/guides/#comp
Exploration of Generative Systems, Journal of ute, accessed 14 February 2021.
the International Association for Shell and
Spatial Structures, December 2012, 53(4): [26] Open Cascade, Open Cascade website,
271. https://www.opencascade.com/, accessed 14
February 2021.
[23] J. Breider, J. Coenders, StructuralComponents:
parametric associative structural design [27] GeometryGym, GeometryGym BIM website,
tools for high rise. In E. Dado, R. Beheshti, https://geometrygym.wordpress.com/bim/,
K.Zreik, Innovations for Building and accessed 14 February 2021.

166