A Real-time Dynamic Decision Approach For Parking Slots Supply In Intelligent Parking System

With the fast growth of private cars in China,parking problem is a major challenge in many metropolitan areas.In addition to scarcity of parking resources,the main reasons to the parking problem are the simple and intuitive management pattern and the lack of scientific operation and management methods.Therefore,they are two essential issues to establish intelligent parking reservation platform with the support of Internet of Things(IoT)technology and develop smart and effective parking reservation management policy and approach based on operation optimization theory.However,it is found that there is still a lack of scientific and reasonable methods for parking reservation management problems based on the investigation of the practical parking management company and the survey.In practice,the simple and intuitive “Firstcome-first-serve(FCFS)” rule is widely used to for parking reservation problem,which cannot achieve the optimal system revenues and the utilization rate of the parking resources due to the random and heterogeneous parking demands.In the literature,dynamic pricing approaches based on revenues management have been proposed for the problem.However,dynamic pricing approaches may not be applied to real-life scenarios,in which the parking price has a very narrow adjustable range and usually regulated by government departments in China.This thesis aims to propose a dynamic supply model and approach of the parking slots for a smart parking reservation platform considering the random and heterogeneous parking demands.The platform real-time decides the number of available parking slots presented in the platform for travelers to reserve to maximize the total revenues of the platform through the finite horizon of each day.We classify the parking demands into different types according to the parking duration and formulate the problem as a multistage stochastic optimization mode considering the reusability of the parking slots during the horizon.To efficiently resolve model,this thesis firstly proposes a data-driven method,i.e.,Two-Sample Average Approximation(2-SAA).Our 2-SAA method uses a half sample data collected from the actual parking management platform to generate a dynamic supply police that determines the numbers of available slots presented in the platform for different type demands showing up in each period.And the 2-SAA method uses the other half sample data to evaluate the generated policy and gives a corresponding the lower bound of the revenues with a predetermined confidence.When it comes to practical application,this thesis applies Rolling-Horizon framework to real-time decide the number of parking slots supplied on the platform based on the current available parking slots in the lot.Finally,we establish an experimental simulation bed based on the real-world parking demands data collected from of “Ebopark” parking management platform in Chengdu of China.The performances of the proposed dynamic supply model and the 2-SAA and Roll-H methods are tested,compared with the FCFS rule and the global optimal approach with the known demands.Besides,we analyze the advantages and disadvantages of the two methods.This thesis addresses the parking reservation problem for smart parking reservation platforms.The proposed dynamic supply model and approaches improves the revenue and the utilization rate of parking slots of parking management platform based on the experimental results.Furthermore,the model and approaches can promote the level of operations management of parking resources in the metropolitan areas in China.