CONTEXT

In the city of Hong Kong today, we can detect multiple problems and challenges. It’s wide-known as an extremely densely populated city, with one of the highest gini index – indicating social and economical inequalities. Spatial segregation between rich and poor that has its roots on the colonial past of Hong Kong. People cannot afford housing in general and are in subhuman conditions of living, the housing crisis gave room to cage homes, shared apartments between many families and roof slums.

On the other hand, the government says it’s impossible to do public housing because all the land it’s already leased to the private developers or its conservation area. The truth is though, that the land is owned by the Chinese government but it’s leased for 50 years to the private developers which do not develop the land because it’s good for them to retain the land and keep rising the prices of the built up area. So eventually, the developers use the excuse to not develop the area because there’s no transportation methods that could absorb more density of people.

TRANSLATION INTO SPATIAL STRATEGY

We can easily proceed into the observation that the transportation accessibility is in direct relation with the paradigms for shaping and structuring of an urban tissue itself.

The objective of our project is to, with the use of Wallacei plugin for Grasshopper, define fixed characteristics of this tissue and use them to provide computational tissue simulations, in order to respond in an intelligent manner to the above mentioned issues.

Hong Kong’s urban tissue and spatial organisation indicate that the presence of a highway which is very wide, influences the ways that urban tissues are shaped on both of it’s sides. We can intuitively understand that not only it is a visible and important separation of two very distinct types of urban fabrics, but also an important factor detaching one society from another in a abstract, yet strong manner. In order to tackle this problematic, we decide to treat both sides of the road as one interrelated mass, one organism living and thriving everywhere in an equitable way. This became the ultimate objective for our exploration. As this city requires denser and denser urban environments, our project takes as a starting point a scenario where the maximum density is obtained. Later on, with an implementation of generative urban tools, the CARVING takes place. Little by little, the optimisation process lets us generate scenarios that are performing the best considering a set objective, such as for example solar gain for the curved mass.

COMPUTING VARIATIONS AND UNDERSTANDING PERFORMANCE

Considering the structure of Wallacei analisis, in order to generate evolutionary designs for the city of Hong Kong 3 main objectives were considered: increase of density (measured by the construction volume), connectivity (measured by the length of the streets of an abstract mobility network) and solar gain (measured by occlusion factors of the shadows in the terrain). It is important to point out how two of the three main objectives have clashing outputs, them being increase of solar gain and increase of density, characteristics which provided the analysis with interesting results.

Setting the 3 main goals, also called fitness criterias in the evolutionary lexic, the Wallacei plugin ran considering 20 individuals throughout 40 generations. This means that 800 different designs for optimizing the urban tissue of Hong Kong were generated, each one evolving in certain aspects.

F01 – density                          ||                       F02 – street length                          ||                       F03 – solar gain

When looking at the results produced by Wallacei, a first effort was made to understand how the exploitation of each fitness criteria could modify the results and create very different geometries. For this purpose, the 3 extreme models were analyzed: one considering the maximum density to be reached, the second one with the highest index of street length and the third one considering the maximum of solar gain in the urban tissue.

Considering these results, it is possible to observe the clashing nature from 2 of the 3 fitness criteria chosen, them being solar gain and density. Looking at the standard deviation graphics and the diamond charts, it becomes clear that whenever there are high indexes for solar gain, density is highly compromised and vice versa. Although, it is also interesting to notice that the solar gain is the most influential factor, also jeopardizing the street length fitness criteria when pulled to the extreme.

CLUSTERING GENERATIONS

Thus, having the understanding of how influential are the fitness criteria and in order to select possible designs which could represent a satisfactory result for all the objectives, it was chosen to analyze the Pareto front¹ results. These results were then clustered by average-linkage².

Within the 4 clusters that were created, one individual was chosen to have the design analyzed. When plotting these options side by side, guided also by the generation of the diamond charts, it was possible to identify one design which had a balanced distribution of the fitness criteria to be reached, which is possible to observe in the image below.

As a conclusion from the analysis, it would be preferable to redefine the calculations of the fitness criteria in a more detailed form and run the analysis considering more generations, since the optimizations results are present in the standard deviation graphics but do not reach a stability yet. This can also be seen in the genome created for each individual generated in the simulation, where it is possible to observe a random genome going towards a more organized one, as generations of design pass.

¹ Pareto efficiency or Pareto optimality is a situation where no individual or preference criterion can be better off without making at least one individual or preference criterion worse off or without any loss thereof.

²  Clustering tries to find structure in data by creating groupings of data with similar characteristics. Average-linkage is where the distance between each pair of observations in each cluster are added up and divided by the number of pairs to get an average inter-cluster distance.

Craving the carving is a project of IAAC, Institute for Advanced Architecture of Catalonia developed at Master in City and Technology in 2021/2022 by Students: Maria Augusta Do Amaral Kroetz, Karim Abillama, Weronika Sojka and Faculty: Milad Showkatbakhsh