Sunlight Access Optimization
Project: Avant Garde · Moscow, Russia
In this study, the client had pre-determined the positioning of five different buildings on an urban Moscow site and was interested in modifying the composition within certain limits to maximize the amount of sunlight that could reach the ground surface. The fixed parameters for the study included the footprint geometries for the five buildings and their locations on the site, the total square footage of development on the site, and values for the minimum and maximum number of stories for each of the five buildings.
For the optimization exercise, BlackBox Studio used a genetic algorithm that was custom developed in-house. Each building had three variables associated with it, corresponding to the x, y, and z coordinates that defined the center point of the roof surfaces, resulting in a genome string with was a total of 15 values in length. The x and y values were constrained to deviate no more than seven meters from the vertical position. The z values were constrained to restrict the towers’ heights from varying only within the allowable limits set by the minimum and maximum floor counts. Because of the constraint that required the development of a specific amount of floor area across the site, the implication was that if one tower increased or decreased in height, another tower would have to adjust in the opposite direction in order to maintain the target square footage.
The GA was run for a total of approximately 300 generations with a population of 75 genomes in each generation, using Ecotect for the incident solar radiation simulations. A number of schemes were visualized from the final generations that all represented well-performing solutions. The example selected for presentation here represents an improvement over the initial design of approximately 7%. There are obviously other performative criteria that would ultimately need to be considered as well (such as cost) and after this project, BlackBox began to look into the use of optimization routines capable of working simultaneously with more than one objective function.
Back to Top