Optimisation vs. Adaptation: Multi-Parameter Optimisation

This analysis of multi-parameter optimisation systems hypothesises their potential roles in the development and implementation of architectural designs. Following a look into how single-parameter optimisation operates in Galapagos, the ‘evolutionary solver’ plug-in for the programme Grasshopper, this post delves into the complex operations Galapagos was created to solve: those that must balance competing objectives. This study also aims to provide insight into another, lesser-known Grasshopper plug-in called Octopus as a means of simplifying and explaining systems of multi-parameter optimisation. A follow-up analysis discusses the capacity for adaptation within these systems of optimisation.

Over the past few decades, facade design and building forms have evolved significantly through the application of environmental simulations and genetic algorithms. Derived from parametric studies of building performance, optimised systems have become commonplace and indicate the growing importance of sustainable considerations in facade design. Environmental performance analytics like Ladybug and Honeybee for Grasshopper have aided immensely in this development.

Facade solutions offer efficient and pragmatic solutions that mitigate harsh climatic conditions affecting buildings and their occupants. Facade optimisation, though calculated for efficiency, fails to address the variety of external conditions, such as sunlight, precipitation and changing occupancy, that affect both a building and its users. Parametric design in architecture can go beyond facade optimisation by combining multi-objective optimisation and efficient adaptability. Systems that achieve this combination can continuously self-optimise by account for fluctuations in performance parameters that affect their ability to compensate for climatic conditions.

Single-Parameter Optimisation

From the onset of the sustainable design movement, designers have favoured optimised design solutions for their ability to distil information and provide efficient, cost-effective and pragmatic solutions for design scenarios. Facade optimisation is a means by which a designer can establish a single design solution that will allow the system to perform best under the full range of exposures it will endure. It comprises static systems that rely on a convergence of data points to highlight the best formation possible based on concrete objectives.

The plug-in Galapagos can analyse only one parameter in a given run, and though that run generates many numeric outputs, it will gradually approach a single best solution. This type of analysis is useful for systems of optimisation as it can test thousands of options and filter through even the most subtle changes in input data.

In this example of single-parameter optimisation, a simple script intends to produce an output with the smallest possible volume. Given three blocks with varying origin points and heights, Grasshopper continuously adjusts the input sliders to slowly create an output that best meets, or optimises, the goal outcome.
In chronological order, these three generations of solutions illustrate Galapagos' ability to iteratively minimise total volume.

In the Galapagos interface, the Options Tab sets the primary goal of the analysis. This could be maximising or minimising a variable, over a certain amount of time, indicated in 'Runtime Limit'.

The Solvers Tab provides a visual representation of the collected data and allows the user to start and stop the analysis. The green bars in the bottom right corner depict each solution by generation. Each can be selected individually and 're-instated', or made to appear, in the Grasshopper workspace.

Under restrictive parameters, optimised facade systems function as a single application. They act as broad-stroke techniques, attempting to provide an effective solution to as much of the problem as possible. These facades, though optimised to meet specific criteria when first designed, fail to adapt to changing environmental conditions or changing user requirements. Once finalised, they are locked into their configurations and any fluctuations in environmental impact need to be compensated for with internal conditioning systems. One way to manage multiple and changing variables is through multi-parameter optimisation.

Multi-Parameter Optimisation

It is impossible to reduce the environmental factors that impact a building to a single parametric relationship. In reality, the factors that affect a building, its envelope and its ability to serve as an effective system of climate control are multi-dimensional. Parametric simulators can aid in understanding variation in environmental performances of a single parameter, but are unable to optimise adequately for all factors in play. Multi-parameter optimisation can achieve median-optimality across a variety of conditions while ensuring a sufficient minimum performance for each condition.

When coordinating a building’s programme or designing its facade, multi-parameter optimisation can act as an instrumental tool to mediate conflicting goals of environmental control, financial costs, client needs and occupant comfort. When designing an effective building system, there is much to take into account, and optimisation in some areas could mean unfortunate compromise in others.

Multi-parameter optimisation is a simple concept. Consider the scenario shown below; a series of blocks of varying heights and base points must produce the minimum possible volume and maximum possible surface area simultaneously. In this case, neither parameter can achieve its maximum potential without hindering the maximum performance of the other objective. Example 1 optimised for minimum volume, but in so doing significantly reduced its maximum surface area; the opposite occurred in Example 3. Through multi-parameter optimisation, a constant cross-referencing of parameters creates a desirable median optimality for all objectives, thus adequately compensating for each performance measure simultaneously. Multi-parameter optimisation can resolve far more than two criteria simultaneously to create an equally dispersed favourability among competing parameters; note the ‘favourable median’ in Examples 2a, 2b and 2c.

Unlike the previous scenario of single-parameter optimisation, multi-parameter optimisation in this example allows for two competing objectives to be optimised proportionally.

While the script and analysis are essentially identical to the prior example, this scenario uses the plug-in Octopus instead of Galapagos. Octopus allows two objectives to be analysed simultaneously.

In the Grasshopper workspace, the Fitness Group segment shows which objectives are under analysis. This example analyses volume and surface area.


Similar to the plug-in Galapagos, Octopus is an evolutionary simulator that can approach optimal solution sets through iterative tests and constant self-adaptation. Unlike Galapagos, however, Octopus possesses the ability to cross-reference multiple parameters simultaneously, whereas Galapagos is limited to a single parametric input. The script below has been broken down into three key components that both inform the analysis and allow the system to run effectively. The pink component to the right is the Octopus plug-in. Similar to the Galapagos plug-in, Octopus requires the same inputs, but as mentioned above, allows the flexibility to input multiple objectives instead of just one. Once Octopus has collected data, it begins to map the information on a coordinate grid that is setup based on the objectives set. Here, one can access the full range of data and separate the solutions that fall in the favourable median from those that do not. After sorting, one can ‘re-instate’ the favourable solutions, or in other words, select their specific data points to appear in Grasshopper, as shown below.

Multi-parameter optimisation with Octopus can generate far more advanced solution sets than simple box scenarios. This script creates a sophisticated analysis to balance the goals of maximising volume and minimising surface area and direct sunlight. With each iteration of the analysis, the script develops a potential building scheme and notes its outcome relative to these three parameters.

Similar to the block example above, the same data trend appears in the tower example below. The towers in Category 1 could be considered the optimal solutions as they adequately balance all three parameters. Note that there is a narrow but distinct range of options amongst these optimal solutions. Each of these solutions falls within the most desirable range of outcomes, but individually possesses its own advantages and disadvantages that would make it more or less favourable for further design development. Categories 2 and 3 represent extremes in the data. Notice in Category 2 how the minimisation of surface area and direct sunlight is ideal, but the volume is compromised greatly. Is multi-parameter optimisation the ideal way to manage changing conditions in parametric design? Read on in the next post, ‘Optimisation vs. Adaptation: Adaptive Facades’.

UNStudio Team: Ryan Henriksen