Optimization in spatial planning from generative design approach: The application for Göktürk, Istanbul

Abstract

Cities, especially with the rapid industrialization period, are facing many problems today. Generating solutions to the needs of cities that are constantly and rapidly changing, unpredictable, influenced by many different systems, and affecting them is challenging for urban planners. While providing fast solutions to the rapidly changing problems of the complex urban system is almost impossible with human power alone, today, with the help of new technological tools such as geographic information systems, artificial intelligence, and computational design, it has become easier for urban planners to manage this chaotic process. These tools assist decision-makers in various stages of the planning process. Still, due to the need to respond quickly to the rapidly changing city, alternative creation stage is often skipped in a long traditional planning process, and decisions are usually made based on a single outcome. Therefore, this thesis aims to provide a framework that facilitates scenario development and alternative selection, often skipped in traditional planning. In addition, this framework aims to enable decision-makers to manage the unpredictable and rapidly changing complexity efficiently. In this study, generative design, one of the subheadings of computational design, was used as the primary work tool. It was selected due to it can focus on the entire process rather than the output product, and it can quickly reach the desired result by making changes within the process when the expected output is not achieved. This feature prevents the urban planner from returning to the planning process's beginning and wasting time. Instead, the urban planner becomes a manager of the whole process rather than the person producing the final product. With these features, the rapidly changing needs of the complex urban system can be responded to quickly within the long conventional planning process, providing time and cost savings. In addition to reacting rapidly, generative design can generate far more alternatives in a short period than human power can handle and helps urban planners make ideal choices among these possibilities. Göktürk neighbourhood of Eyüpsultan county, Istanbul, was selected to implement this tool as the case study area. Göktürk, where demand is increasing rapidly, has both gated communities and rural characteristics. Although, at first, it may seem that the gated communities in this area provide adequate facilities, they do not respond to the needs of the rapidly changing city due to the privatization of services. Therefore, Göktürk, chosen due to its complexity, was first analyzed to see how much it meets the demands. In the first part, a population projection was made for the year 2041, and it was investigated how adequate it will be in 2041, assuming the current land use has not changed at all. The Spatial Plan Legislation (2014) determines the minimum area requirements and walking distances that were also the limit for the analysis. The results show that all the kindergarten, primary school, secondary school, high school, health areas, sociocultural areas and worship areas are below the standards regarding the area size. In addition, accessibility ratios are also lower than the ideal, except for high school (98,9%). For the second part, the same analyses were made for plan decisions, and it was seen that land use decisions don't meet the standards like in the current situation. In the second part of the case study, three basic scenarios were primarily determined to support the urban planner's decision-making process for Göktürk. The first scenario was made to fill the gaps in the current situation; the second was also made to fill the gaps between the ideal situation and plan decisions; the third scenario was made to see the area's full potential, assuming that there was no defined land-use decision before. For this purpose, the Rhino Grasshopper plug-in was used as a generative design interface to create the design algorithm. For the algorithm, random parcel selection within limits determined for kindergarten, primary school, secondary school, high school, children's playground, health area and religious areas were parametrically introduced. These aims defined 14 fitness objectives of the algorithm as minimizing the difference between ideal and existing area size and maximizing the accessible housing parcels for kindergarten, primary school, secondary school, high school, children's playground, health area and religious area. In total, 15000 solutions were generated and after, these solutions were optimized by using NSGA-2(non-sorted genetic algorithm 2), 2835 of them located on the Pareto front. Finally, to narrow the selection space, the fittest solutions of each fitness objective and the average fitness rank = 0 of each scenario were represented to the decision-maker. To sum up, this new framework attempted to adapt to the rapidly changing city's need and control the planning process with multiple alternatives within the limits of Turkish planning standards for areal adequacy and accessibility. It is believed that this tool will start further discussions about a collaboration of urban planning and new technological decision support tools, their implementation at municipal levels, and their contributions to participation in decision-making by representing the alternatives and helping to select via optimizing.