Computational Design and Machine Learning in MEP Systems for Large-Scale Architecture
Interview with Gabriel Garcia of Royal HaskoningDHV
Gabriel Garcia of Royal HaskoningDHV ‘s work blends computational design and the foundational principles of architectural engineering.
His expertise has been pivotal in enhancing the efficiency of projects for notable clients, including the collaborative work with NACO for airport design and contributions to projects led by Foster + Partners, and the Bjarke Ingels Group. Within the Buildings Department at RHDHV, Gabriel crafts custom software solutions that streamline the complex processes of structural and installation design, bridging the gap between the architects’ visions and engineering realities.
This conversation with Gabriel sheds light on how computational design principles are applied in the execution of large-scale projects such as airports.
He offers insights into the process of integrating computational strategies to optimize designs and solve complex challenges. From data cabling systems in airports and the assessment of sun reflections on roof panels, Gabriel illustrates the practical application of his work.
As we explore Gabriel’s contributions to Royal HaskoningDHV and the broader architectural community, we touch on how his approach is not just about the buildings themselves but also about how these projects mirror the complex systems found in cities and even the human brain.
Computational design’s ability to transcend scale, offering a glimpse into the potential for these principles to shape the future of not only architecture but also the interconnected systems that define our world.
As a Computational Design Specialist at Royal HaskoningDHV, can you give an overview of your role within the company?
I work in the Buildings department at Royal HaskoningDHV. My role involves enhancing and streamlining traditional engineering processes, such as structural and installation design, building physics analyses, and addressing technical architectural challenges.
In my daily work, I develop small programs to optimize the efficiency of engineers and architects. I often describe my role as that of a software prototyper, as I create custom software tailored to specific tasks in building design.
How do you integrate computational design principles into large-scale building projects, and what does a typical project lifecycle look like for you?
Nowadays, data plays a crucial role in building design. Every aspect of a modern building is shaped by a combination of human expertise and design rules.
While individual human knowledge can be challenging to replicate computationally, design rules can be translated into computer algorithms.
The complexity of these design rules determines the effort required to create the corresponding algorithms, which can sometimes be time-consuming and costly. However, when the process behind these algorithms is repeatedly applied—especially in large-scale building projects—it becomes worthwhile.
In our company, we initiate automation processes when individuals or teams spend significant time on repetitive or tedious tasks.
I firmly believe that humans should focus on creative aspects, leaving the repetitive work to computers. At this juncture, I develop software prototypes tailored for these repetitive tasks. If the prototype proves useful and is utilized frequently, it can then evolve into a fully-fledged software solution.
Computational Team Dynamics and Contributions
Could you describe the composition and focus of your computational design team?
In my team, we have six computational designers. Typically, we collaborate in pairs on each project, and once a month, we come together to share experiences and knowledge.
How does your team function within the broader context of Royal HaskoningDHV, and what role does the team play in the company’s architectural and engineering projects?
Within the Buildings section of RHDHV, we have several teams similar to mine. Each team specializes in specific areas such as structural design, building sustainability, architecture, and more. In addition to the building department, RHDHV also operates other business lines, including Water & Maritime, Mobility & Infrastructure, among others. Each of these business lines has its own computational designers.
As a computational designer, it is common to collaborate on projects with colleagues from other business lines.
Our company gives a huge importance to the community of computational designers who gathers every 2 or 3 months to share insights about new technologies and projects developed using computational design.
Project examples and process
Can you share some examples of projects your team has worked on that you will be presenting at CDFAM, and how have you employed computational design strategies to solve complex problems, optimize designs, or enhance the functionality of these projects?
In the CDFAM, I will be presenting the collaborative work I undertook with the aviation department of our company, NACO. Airports are one of the most complex building typologies I ever worked on.
In a forthcoming airport project, Foster and Partners served as the primary project architects. During the architectural design phase, they provided us with weekly updates, occasionally introducing radical changes to the floorplans.
At NACO, specifically within the telecommunications department, our task was to design the data cabling system and determine the optimal placement of the technical rooms, in this case, approximately 160. These rooms varied in size from 6 to 20 square meters, depending on the amount of data outlets connected to it. The primary challenge lay in ensuring that data cables did not exceed a maximum length of 90 meters, while also managing the maximum number of data outlets connected to the servers within each technical room. Additionally, the positioning of data outlets was driven by the ever-changing floorplan layout, and to generate a new floorplan layout, the architect needed our feedback. So, a weekly basis loop.
In order to achieve this, during the early design stages or when significant floorplan modifications occurred, we employed a machine learning algorithm to efficiently position the technical rooms while adhering to all the above mentioned constraints. For minor adjustments or small design tests, the tool could be used in manual mode, allowing for fine-tuning.
In another airport project, this time led by the Bjarke Ingels Group, our task was assessing whether sun reflections on roof panels could affect landing airplanes.
The challenges were: a large number of panels on the roof, multiple bounce reflections in neighbor panels, multiple runways, and the huge amount of calculations—specifically for this project, one calculation for every minute of the year. To address this, we adopted a geometrical approach, calculating millions of reflections.
Software Tools and In-House Developments
What commercial design and engineering software tools are typically used, and how do you integrate these with custom solutions you and your teams develop in-house?
At RHDHV, we have a variety of in-house tools tailored for different purposes. Each discipline has their own toolkits. However, when starting a new project that involves design exploration or interactive solutions, we often use Grasshopper and Rhino. If a custom tool is required, we can create a Grasshopper plug-in to seamlessly integrate it into our design process.
Additionally, we frequently utilize Speckle, a commercial software that serves as an open-source platform for handling 3D data. Whether for data exchange or plug-in development, Speckle proves to be a valuable resource.
Could you discuss any gaps in commercial software that your team addresses through proprietary tools or plugins, and how these help to realise your projects?
In our company, one of the most widely used self-made plug-ins is the Open Data Toolkit. This toolkit comprises a suite of tools designed to collect spatial data from open sources and seamlessly integrate it into Grasshopper workflows.
An illustrative example of the data we work with is the 3D BAG, which represents 3D buildings across the Netherlands or the AHN, which is the digital terrain model of the whole country. Additionally, we harness data related to trees, roads, and other urban features available from open governmental sources.
While our toolkit performs well with Dutch data, we are actively expanding its capabilities to cover other European countries. We view this software as a vital bridge between the worlds of GIS (Geographic Information Systems) and CAD/BIM (Computer-Aided Design/Building Information Modeling)—a critical connection that remains underserved by commercial software solutions today.
Code development and distribution process
How does your team approach the development of new code, particularly with the intent of sharing these tools both internally and externally, such as on platforms like Food4Rhino
At RHDHV, we have our own distribution platform where anyone in the company can share new scripts along with their respective instructions for use. Within this platform, we host a collection of tools that are both fully developed and under ongoing refinement.
Additionally, we are currently exploring the possibility of sharing some of our tools on Food4Rhino, although this initiative is still in the discussion phase.
Balancing multi objective design
Given the complexity of projects you are working on, how do you gather and rank the data to balance the multi-objective criteria in your projects?
In multi-objective optimization, we often employ tools similar to the well-known Design Explorer. With such tools, we can present results interactively to clients or stakeholders, especially because they have a key role in the decision-making.
Can you envision applications of these same principles in fields beyond architecture, such as bioprinting for organs, where the relationship from parts to whole must be addressed from the cellular level?
While architecture and buildings often lead the way in many aspects of computational design, the underlying principles can be applied to any discipline.
My background lies in architecture, but I also studied technology at both city and regional scales. I learned how to create vector maps using satellite imagery, reading related to soil humidity and temperature. Additionally, I used my skills in parsing data from entire cities, collecting its essence and transforming it into the rules to create a new residential neighborhood.
Upon joining RHDHV, I began applying shortest path algorithms within buildings, these are typically used for road networks and route calculations (similar to Google Maps). These algorithms helped me find the most efficient paths for ventilation ducts and data cables.
The key lies in abstracting knowledge and adapting the scale of the challenge.
I often drew parallels: comparing a city’s road network to a building’s intricate system of pipes and cables, or even likening it to the human body with its veins and nerves.
Taking this analogy further, consider the human brain with its complex neural network and add it as a fourth system in the comparison. Remarkably, all four systems—cities, buildings, bodies, and brains—share a common thread: these have sensors that collect data and generate responses.
Takeaways from CDFAM
What do you hope other participants take away from your presentation at CDFAM? Are there specific insights or trends in computational design you believe are essential for the future of architecture and engineering?
I don’t know, maybe that idea of knowledge abstraction. I mean, the notion that computational design thinking transcends scale, regardless of how one acquired it, is indeed powerful. I briefly explained this concept in the previous question.
The knowledge gained during the CDFAM becomes a versatile toolkit that can be used effectively in any context. Whether we’re designing buildings, shaping objects or unraveling the complexities of biological systems, this adaptable mindset empowers us to create meaningful solutions.
Lastly, what are you looking to learn or gain from your involvement in the CDFAM Symposium? Are there areas within computational design or collaborative opportunities you’re particularly interested in exploring further?
I am always open to learn from different contexts. I appreciate that the CDFAM isn’t exclusively centered around buildings. In the past, I’ve employed computational design for creating jewelry, furniture, and 3D printing, among other things. Unconsciously (or consciously), these diverse experiences have contributed to my current ability to apply computational design knowledge for creating better buildings.
