data jungle adventures

Generative Design (GD), or “the future of making things”, promises a computerised way to answer design questions through specifying input, goals, and constraints. An effective co-designer for the human mind. Or maybe more?

What is generative design today?

By definition, generative design is a Computer-Aided Design (CAD) engineering process through which a designer collaborates with artificial intelligence algorithms to generate and evaluate hundreds of potential designs for a product idea.

Traditionally, the design process includes several design generations, known as iterations, and even sometimes it requires starting all over again. Algorithms iteratively “create and stimulate thousands of designs in a fraction of the time”, much faster than any design expert. The time efficiency, flexibility and possibility to set quantifiable constraints (material thickness, travelling times, material properties, etc) along with the broad range of design scenarios are some strong points of the generative design process.

As described by Siemens, the design process starts with setting goals and constraints. Examples of those design parameters can involve the following:

• Physical properties: product size, geometric dimensions, weight, material properties, manufacturing methods
• Material responses or behaviour: permissible loads and operating conditions of used materials
• Financial factors: cost per unit.

The process though is made up of seven steps: defining the problem, gathering the data, setting the evaluation criteria, generating the model, evaluating the results, evolving the design, and selecting and refining.

In addition, the combination of generative design with modern forms of manufacturing, like 3D printing, solves the problem of accessibility to manufacturing for startups and small businesses and reduces consumerism and contributes to sustainability in design.

Can generative design be applied in landscape planning?

Amid the relative scarcity in the literature on the application of generative design for landscape architecture, we choose a few references that give us a glimpse of the role of data, software, and Generative Landscape Design Systems in the application of GD in landscape planning.

How important is data?

In the application of GD for landscape projects, we need to talk about data and its elementary role. In this context, one of the use cases studies the generative modelling approach for open space design. The process of modelling was defined in four algorithmically associated stages.

• Field observations for data collection of vegetation and user behaviours for site digitalization,
• Algorithmic parameters defined
• Design constraints identified
• Evaluation of all parameters with constraints.

No data, no design. This study emphasizes the importance of the data for an open space design more than the final design solution though. The input drives and controls the output. Commonly known as the “rubbish in rubbish out” rule.

Generative landscapes design system…

Charalampidis and Tsalikidis, on the parametric landscape design approach for urban green infrastructure development, mention that the Generative Landscape Design System (GLDS) uses mathematical models of physical processes, algorithms, and statistical analysis in order to increase the environmental and ecological performance of a certain design solution. The GLDS used consists of four subsystems aiming at localising planting zones. The first generates the terrain. The second organizes the space. The third creates the shading structure and the fourth defines the planting zones. Necessary initial input data is required for weather, topography, orientation, sunlight obstacles, and plot boundaries.

This study results in a master plan where planting zones are plotted along with lawn, benches, playground, walkways, main movements, resting areas, existing trees, proposed trees, main entrances, and secondary entrances.  

Software for generative design 

Generative design tools are either CAD systems with added functionality through modules (Fusion 360, Solid Edge, Creo 7.0, CATIA V6, NX, ) or standalone software solutions with data exchange feature to talk to different CAD systems (CogniCAD, and MSC Apex). Most of the software is commonly used for civil or mechanical engineering tasks. 

For us, the main tool to be used for GD in Landscape architecture is grasshopper, the graphical algorithm editor in the Rhino family. Unlike RhinoScript and RhinoPython, grasshopper requires no knowledge of programming or scripting. Frequently a combination of tools is required to answer complex real-world questions in urban and landscape planning. In Charalampidis and Tsalikidis’ study, they used a combination of Rhinoceros 3D modeller, Grasshopper, and Ecotect for performance analysis using weather data, Geco (a set of components to link Rhinoceros/Grasshopper and Ecotect), Kangaroo (a live physics engine), and Galapagos (problem-solving engine). 

On another side, the blog generative landscapes show a remarkable exploration of algorithmic systems and their potential application in landscape design and architecture. The blog describes specific script examples in Grasshopper for various subjects: random scaling of circles, complex patterns from simple arcs, multiple point attractors with a random point cloud, surfaces from an image sampler, and many others. The next blog post will explore these and much more!

Conclusion

The stages of generative design require human input at several steps. However, bringing AI into a design process gives the impression that the competition between the human mind and machine is becoming harder. 

Clearly, there is no superpower evolving generative design. It is a contest between the human and himself, his yesterday and his tomorrow.

For now, let’s limit the comparison between generative design and human creativity to two points. A person without aims in life is like a generative design process without goals set. And a roaming person with no moral behavioural barriers is like a generative design process with no constraints…