A GA based approach to location selection and dimensioning of automated parking facilities

Number of vehicles increasing directly proportional to rapid population growth has also led transportation problems to increase in many cities of the world. Parking problem is one of the most important transportation issues in today's metropolises. The fact that parking spaces are inadequate despite the rapid increase in the number of vehicles and all the negative effects at the urban scale caused by this fact can be covered within the scope of "parking problem". An average car spends 23 hours a day, i.e. 95 percent of its lifetime, parked. This period of time is much greater than the time period in which the vehicle is in motion. As it is necessary to supply roads for the movement of vehicles, sufficient parking space is needed for them to be parked. According to international literature, there are 3 different approaches to parking policy in the world. These can be categorized as conventional approach, parking management approach and market-based approach. Conventional parking policy is implemented in most of the world's cities. In this approach, parking problem is perceived as an infrastructure issue. The main objective of this approach is to meet the demand for parking. Minimum parking requirements are the primary regulation tool of the conventional approach. According to conventional parking policy, each building must provide enough parking space to meet its own anticipated maximum parking demand. Since conventional approach defines parking problem as "lack of parking space", it recommends each building to meet its own demand in its own building plot as the solution. Minimum parking requirements are determined depending to kind of land use and parking regulations are prepared within the scope of the proposed solution. According to international literature; conventional approach, which is based on minimum parking requirements, is incompatible with problematic and dense urban fabric. Parking management is the second most common approach among parking policies. This approach has emerged due to the fact that conventional parking policy is not feasible for dense and congested areas. Parking management approach is usually implemented for the areas where parking problem exists but it is not possible, or desirable, to provide extra parking supply. Parking management consists of policies and programs aiming more efficient use parking spaces. In market-based parking policy, balance between parking supply and demand is expected to be achieved through market-based pricing. Spillover is not seen as a negative issue in the market-based approach, but it is seen as a state that is keeping local parking market alive and can be prevented through pricing. Steps towards balancing parking supply and demand should be taken to solve the parking problem in Istanbul. In the current situation it is known that parking demand is much higher than parking supply and the consequences of this fact are evident. Methods to be applied in balancing parking supply and demand can be based on reducing demand or increasing supply. Only increasing supply is not a sustainable solution itself. If supply is increased only, it can be anticipated that each parking space will create its own demand and car-dependent transportation will be encouraged. In an ideal solution, both current parking demand should be met and steps towards minimizing probable demand in future should be taken. This thesis focuses on the steps towards increasing supply. It is possible to increase supply only by creating new parking spaces. Accordingly, new parking facilities should be planned and built by taking current status of Istanbul (high population density, unplanned urbanization, etc.) into consideration. The planning process of parking facilities to be newly built is comprised of three stages. These are: selection of parking facility type, selection of parking facility location and dimensioning of parking facility. A novel GA-based computational model which is supposed to support decision-maker/designer in site selection and dimensioning stages of parking facility planning is presented in this thesis. After, introduced model is tested with real-world data in the context of a case study and successful results are achieved. Types of parking and parking facilities are discussed while deciding the type of facility. There are two types of parking regarding parking space's relationship with street: on-street and off-street. Off-street parking can be analyzed in two groups of ground-level or multi-storey. Multi-storey parking facilities are divided into static or automatic relative to how vehicles are parked. Cars are parked by drivers in static facilities. In automatic facilities, parking is accomplished through lifts that are able to move horizontally and vertically in order to carry and stack vehicles into parking cells. It is not always possible to find a land to build a new parking facility in cities like Istanbul which have high population density and unplanned urbanization problem. Available lands are usually too small to build a large structures on or too expensive given cost of alternative investment. In this thesis it is concluded that automatic parking facilities are the most appropriate parking systems for Istanbul case, since they allow maximum number of vehicles to be parked in the minimum amount of space. At least one criterion such as cost, profit, travel distance, service, waiting time, coverage, market share, etc. is aimed to be optimized while determining facility location. In the section about facility location; a literature review on location science is presented, facility location problems are classified and solution procedures used in parking facility location problems are discussed. In the section about dimensioning of facilities, the factors determining the dimensions of a parking facility and the methods to calculate parking demand are covered. In the facility planning phase, it is necessary to know the local parking demand and supply in order to determine the capacity of the new parking facility to be built. This data is obtained through parking surveys. Parking surveys are made via methods such as survey with drivers, count on field, count on photos obtained by satellite or airplane and count on GIS. Surrounding urban fabric is another factor to be considered when determining facility dimensions. Size and scale of parking facilities should be compatible with surrounding buildings. Genetic algorithms, linear programming, weighted sum method and penalty functions are the methods used in creating the model that is introduced in this thesis. Genetic Algorithm (GA), which is inspired by Darwin's theory of evolution, is a solution procedure frequently used in optimization problems. Linear programming (LP) is a method used in order to optimize allocation of limited resources. A typical LP problem consists of three components: objective function, problem constraints and non-negative variables. Weighted sum method is a method for converting a multi-objective optimization problem to a single-objective optimization problem. In this method, each objective function is multiplied by scalar weights determined by decision maker and all the weighted objective functions are summed up. Thus the problem can be expressed by one function, and can now be solved by any single-objective optimization method. Due to its nature, GA can only be used in unconstrained optimization problems. Constrained optimization problems can be converted into unconstrained optimization problems using penalty functions. Thus, it becomes possible to use GA also for constrained optimization problems. By this model; two different multi-criteria problems such as facility location and dimensioning, which are both NP-hard, are solved simultaneously. The model is created using Rhinoceros, a 3-dimensional modeling software, and Grasshopper which is a visual programming environment integrated with Rhinoceros. The graphical data of the land in Rhinoceros is associated with the algorithm that is created in Grasshopper. In addition; Galapagos, which is a component in Grasshopper, is utilised to be able work with GA. There are many conflicting criteria in this model. These criteria can be grouped under four main headings: the cost criteria, the accessibility criterion, the social criterion and the technical criteria. Şişli, which is one of the most problematic districts in Istanbul in terms of parking, is selected for the case study. Study area is an area of 18 870 m2 which is located in Şişli District, Mahmut Şevket Paşa Neighborhood. This area consists of 81 buildings which are mostly residential. Since there is not enough data on commercial buildings in this area, only residential parking demand was covered. The parking facilities to be generated are planned to meet the existing residential parking demand. In the context of this case study, two different scenarios are generated using different GA parameters. One successful solution is achieved in each scenario. By this case study, success of GA in problems such as facility location and dimensioning is proven once again.