Abstract

The world’s building stock is expected to double by 2060, which means building one New York City every month over the next 40 years. In order to achieve this in a sustainable way, we need smarter ways of designing our cities. Spacemaker is strongly committed to contributing to this purpose by providing real estate developers and architects with a software that allows for immediate optimisation and analysis of any building site. However, the site development problem entails navigating a high-dimensional solution space where the system needs to take into account all buildings, surroundings and other functions on a site. To make it more challenging, our work has shown that it is infeasible to capture all requirements and preferences of the user in an explicit objective function. To give our users a tool that allows them to navigate and explore this vast solution space, Spacemaker has developed a human-in-the-loop system for optimization that takes advantage of the superior problem-solving capacity of a hybrid human-machine system. In order to accomplish this, we use a wide range of techniques including search/optimisation, sampling (GANs etc.), dimensionality reduction and visualisation techniques. This talk will introduce you to some of them, both generally and in the Spacemaker use-case.