The breeding criteria was pre-defined and the sequence diagrams were carried out computationally. All three generations were evaluated in the end. In this case only three consecutive generations were ran; in actual design projects, the number of generations ran would be much higher.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saad

The chain of parallel and successive operations build complexity. A breeding strategy was set with the intent to evaluate the fitness and form variation of the populations by applying different parameters for mutation probability, crossover rate and parents.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saad

The logic for defining the crossover strategy relied on a 50-50 percentage of genes for breeding between fit individuals to a gradual 90-10 percentage of genes for breeding between fit and unfit individuals. The aim was to obtain a very fit population.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saad

The genes and objectives inform the Octopus Genetic Algorithm that, through computational processes, would create generations composed of individuals with different phenotypes based on the sequence, magnitude and types of operations that are applied on the different sub-divisions of the body plan. The genetic algorithm aims to fulfil the objectives defined by manipulating these parameters.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saad

The aim of the seminar is to employ the logics abstracted from biological evolution within a genetic algorithm for the evolution of geometric form in architecture. The sexual reproduction is simulated through breeding and gene-crossover techniques. Mutation in the structure of the genome is also introduced in order to further develop form.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saade

Through these breeding strategies, increasing the mutation rate contributes in obtaining a more spread population which may or may not create hopeful monsters to lead the next generation towards better fitness. On the other hand, increasing elitism, that is, increasing the percentage of breeding between fitter parents would only give a fitter population.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saad

In an attempt to increase the fitness variation, the next generation was ran increasing the mutation probability. The result was that buildings radiation was increased instead of courtyards radiation. The interpretation of such result relies on the existence of competing criteria. In some cases one criteria gets satisfied while compromising on the performance of the other.

Georgios Bitsianis
Ernesto Pastore
Abhilasha Porwal
Camille Saad

Experiments of emergence began on a primative geometry - the sphere. The individuals were ranked based on a fitness criteria. An evolutionary goal was set for the population and further generations were bred in order to achieve this goal. The logic behind the experiment remained the same during each generation but the complexity was increased by introducing a body plan for the primitive.

Radhika Amin
Samidha Kowli
Stanley Carroll
Zeynep Aksoz

For the mutation strategies, deletion and inversion techniques were implemented.

All the individuals from the first and second generation were ranked together according to the fitness criteria and the individual, highest in the ranking, was bred with each of the next strongest four individuals in the ranking.

Radhika Amin
Samidha Kowli
Stanley Carroll
Zeynep Aksoz

The flow of the algorithm (the grasshopper definition) played a crucial role in determining which genes led to the morphological changes in the population. In the GH definition, the genes were set as the parameters of street offset, courtyard widths and the distance from the open edge of the plot. The gene for number of floors per division of the body plan and their scaling, acted like the HOX (HOMEOBOX) genes for the overall morphology, differentiation and variation in the population.

Radhika Amin
Samidha Kowli
Stanley Carroll
Zeynep Aksoz

Unlike the traditional methods of algorithms which are linear processes initiating from one point and end at a sub-optimal solution, Genetic Algorithms deal with a set (a population) of possible solutions (individuals) in relation to a multi-objective formula. In order to develop a further understanding of the Octopus plug-in of Grasshopper and the effects of multi objective optimization on building morphology over many generations, a single urban block was studied and 30 generations were ran with different tests to evaluate the emerging morphological patterns.

Radhika Amin
Samidha Kowli
Stanley Carroll
Zeynep Aksoz

Evo-Devo (Evolutionary developmental biology) draws a relationship between the development of forms and the evolution of their developmental processes over time. The experiment was the evolution of an urban block using Octopus, an evolutionary optimization plug-in for grasshopper in Rhinoceros, to understand and evaluate the advantages and limitations of this process with reference to an urban context.

Daniel Zaldivar
Suhash Patel
Silvia Daurelio
Amritha Krishnan

The evaluation criteria for building morphology is the maximization of density with a high floor area ratio. Because this criterion competes against the necessity for increased semi-public space, the area of the courtyard with solar exposure at noon is set up to be maximized during winter and minimized in summer; these criteria are mutually exclusive.

Daniel Zaldivar
Suhash Patel
Silvia Daurelio
Amritha Krishnan

Through a combination of genes, the genetic algorithm found a specific pattern where the elimination of sections on the south facade of the building permit a high winter sun-vector exposure while having the courtyard covered from above. Computationally, it favours all of the fitness criteria within one solution.

Daniel Zaldivar
Suhash Patel
Silvia Daurelio
Amritha Krishnan

A typical computational set-up for Octopus relies heavily on a vast number of homeobox genes: in many cases, the need for differential behaviour of different parts of the body plan necessitates different genes to encode the same function in each. With computational tools and speed, large amounts of data can be calculated simultaneously. Because of the non-linear nature of problems in engineering and architecture, there are lots of variables and interacting subsystems. When these forces and criteria governing them can be quantified, computational power can then be exploited to generate solutions in an expedited manner.

Daniel Zaldivar
Suhash Patel
Silvia Daurelio
Amritha Krishnan

Genetic Algorithms can be used as an optimization tool to search the design space for solutions and to explore options and possibilities in the early design stages.

Shahad Abdulmonem
Faisal Al Barazi
Jose Garcia
Alessandra Lazzoni

Key to understanding natural forms is their development in evolution. Evolutionary development, through the interaction of simple rules stemming from a single cell complex, allows for a diverse and natural system to emerge.

As architecture and the built environment is a non linear complex system of interconnected forces, a great deal can be learned from the Evo Devo to explore and search the design space for a variety of possibilities and are not bound by the pre-conceived image of the built environment.

Shahad Abdulmonem
Faisal Al Barazi
Jose Garcia
Alessandra Lazzoni