Car Park Overspill Study PROJECT OVERVIEW
Contents
Project Introduction and Background
Together with the progressive development of the society, density of urban population has increased and cities' infrastructure have become more and more complex. In the aspect of transportation, one of the crucial issue is the management of urban traffic growth.
While Singapore is a small country which has limited land space, in order to utilize the land and improve the city infrastructure, the Land Transport Authority(LTA) aims to do an analysis to understand the current parking situation at these selected locations. The LTA Contract Parking Study was awarded to Media Research Consultants in March 2015 to undertake the Study involving 65 car park locations in Singapore, including 30 retail malls, 15 retail and F&B clusters in landed housing estates, 10 hawker centres, and 10 community clubs.
This study was to conduct parking occupancy surveys, human traffic counts, and interview surveys at the selected locations at stipulated times. The study incorporated the conventional method of manual counting as well as deployment of automated counting equipment. Face-to-face interviews were employed for the interview survey segment.
In addition to the study and taking of detailed data from field surveys, this project aims to develop a simulation tool that enables the systematic analysis of the impacts of various parameters, using a collected set of quantifiable data.
Motivation & Objectives
The limited land space in Singapore is a challenge that is here to stay hence the best solution is to utilise every single space more effectively and efficiently; and within statutory regulations. This study will focus on facilitating the future design of car parks in Singapore by maximising its effectiveness of the land space required. This requires a careful analysis and evaluation of the impacts from different factors.
Business objective
Analyzing of data by identifying and describing the current trends of the parking situation in various locations. To build a model which is able to simulate the car park operation with different parameter inputs so that the relevant authorities is able to forecast the optimal car park size required according to the various factors in the surrounding environment. Hence to increase the land/car park utilisation rate.
Technical objective
To apply basic analytical skills through data cleaning, data standardization, data process to identify the current trends of car parking situation and the properties of each individual as well as the environment. This will help us develop a deeper understanding in applying analytical skills in a real world scenario.
To learn to use a computational modelling technique such as agent-based modelling to simulate and recreate a complex phenomenon where each autonomous individual follows their own rules and properties to act and interact with the surroundings. This approach help us to gain higher level information about parking situation by simulating the what-if scenario against different parameters such as car park sizes, car park associated site categories, and whether it is peak or non-peak day etc.
Project Scope
With data sets provided by our project sponsor, our group will help our sponsor conduct some preliminary data analysis in report format. The preliminary report contains detailed site information along with relevant data collected, such as human traffic count within the site and roadside parking data. We help our project sponsor to compile the reports so that our sponsor can further hand them over to Land Transport Authority (LTA) to let LTA have a better understanding about the parking situation around those identified development sites. In this project, our group aims to finish 5 to 6 preliminary reports, and each of them is for one particular site that assigned by the project sponsor.
Meanwhile, having identified the limitations of the collected data provided by our project sponsor, in order to conduct a valid data analysis and gain more insights, our group aims to build a simulation programme based on Agent-based Model to simulate and estimate car park parking occupancy and overspill outcomes against a set of dynamically changed variables, and to provide a valid explanation on the outcomes of our model against the variables in report format. In order to achieve these objectives, our group identified the following major tasks in our project besides the preliminary reports aforementioned:
- Prepare data including cleaning collected raw data and transforming them into usable format.
- Research on Agent-based Modelling (ABM) and learn the programming language required by NetLogo - an ABM simulator.
- Apply project data into NetLogo and build a valid model.
- Analyze and evaluate the outcomes of our built model.
Project Data
Data Collection
In order to facilitate the understanding of current parking situation, two kinds of data is collected in this project by our project sponsor MRC. The data covers information about Parking Occupancy, Parking Overspill and Human Traffic Count and they are collected island wide. In total, there are 65 different site data being collected, and the sites include 30 retail malls, 15 retail and F&B clusters, 10 hawker centers, and 10 community clubs.
Parking Occupancy and Overspill data counts the total number of car park lots available on site, off-street parking lots provided by the development, and the number of parking lots inside the public car parks. To collect the data, 3 to 4 Enumerators are assigned to the entrance / exit of the car park. A typical example would be, Enumerator A counts the number of empty lots in the park, Enumerator B notes the number of cars IN and OUT until the intended start of the survey. Hence, the following equation si derived: the Base Car Park Occupancy (the number of empty lots in the park at the survey time) = number of empty lots in the park - Total IN + Total OUT.
Human Traffic Counts refers to the counting of the number of people that enter and exit a particular development, in order to derive the total number of people within the development at a particular period. Five methods are used for counting, and they are:
- Manual Count
- Line-of-Sight Count by enumerators
- Automated Count using electronic devices
- Snapshot Count
- Sit-Stand Count by roamers
In considering the possible differences of data collected under Peak and Non-peak days, aforementioned data is collected in week days and weekends separately.
In order to assure the survey sample validity, the margin of error and confidence interval of the sample data is controlled within 5% and 95% respectively.
Data Set
Pre-survey Report
It states the basic information about the survey site, such as site characteristics, site assessment, recommended survey time and locations. The report is to provide readers more insights into the studied area and assess the impact on the surroundings.
Human Traffic Count
It records the total counts of human in and out in 15 minutes interval from 10:00 AM to 01:00 PM and 6:00 PM to 9 PM each survey day.
Roadside Parking Count
It records the car park occupancy and overspill count in 15 minutes interval from 07:00 AM to 10:00 PM of each survey day. The data is further subcategorised into Cars and M/Cs.
Interview Survey
The survey states some demographic information like genders, age groups etc along with other information like the purpose of visit of surveyors, kinds of transportations taken, visit frequency etc.
Final Deliverables
Phase 1
- Project proposal
- 5 to 6 preliminary analysis reports for MRC and LTA to review.
Phase 2
- Midterm presentation.
- Project poster
- A programme built on agent-based model that simulates a car park overspill performance against a set of dynamically changed #variables representing site characteristics.
- A report analyzes and evaluates our built model and states the model’s assumptions and limitations.
- Final presentation & demo.
Methodolgy
After understanding and analyzing the nature of the LTA project, we feel that the data collected can be used for more thorough studying of potential relationships between carpark size and sites’ nature and size. Moreover, due to the fact that the utilization of the carpark depends on every individual drivers’ behavior, it is almost impossible to simulate the problem on a Macro level since we cannot coordinate and predict everyone’s behavior in the big environment. Therefore, the approach from a Micro level is more applicable in this case since we can simply defined individual’s behavior and let them interact freely to simulate the problem and generate the results. Hence, we have decided to apply Agent-based Simulation to the carpark issue.
Agent-based Model
Agent-based Simulation is a microscale model that simulates the simultaneous operations and interactions of multiple agents in an attempt to re-create and predict the appearance of complex phenomena. In our case, the agents are the people who travel to the sites and the drivers who park their cars. By defining individual agent’s behavior based on the data collected, we can simplified the issue. For example, a single rational driver’s behavior can be defined into looking for a slot in the carpark when he travels to the site, if there aren’t any slots, he can choose to park illegally on the street or leave for another site. Such behavior and the probabilities of making each decision can be generated from the historical data. When we have a model which contains thousands of drivers who are rational and a carpark in certain size, their interactions will tell us whether the carpark has enough slots and how big is the overspill. This model can be then used for future planning of new sites and carpark to test whether the allocation of carpark slots is efficient. Based on this design of the Agent-based Simulation, we need to split the project into 2 Phases:
- Analyzing the data to derive agent behavior
- Build Agent-based Model using the behavior derived in Phase 1
Phase 1:
1. Data cleaning and grouping
The carpark usage data can be naturally grouped into 4 categories based on the sites’ nature: Shopping Mall, Community Centre, Hawker Centre and F&B Clusters. If we assume that the agents in each category will share similar behavior, we can then analyze the carpark usage data by category to understand the agent behavior better.
2. Defining agent behavior
By studying the data by category, we can find out the distributions of the amount of cars which will travel to the sites and use the carpark. Also the chances that they may park on the street illegally. More complex behavior can be investigated if time allows.
Phrase 2:
1. Build Agent-based Model
Developing Agent-based Model using Netlogo to simulate the carpark problem. The basic setting will include the site, the people, the carpark and the cars. The code will need to define each components’ behavior correctly, for example, the cars will try to find a slot in carpark before the people in the car can get off and travel to the site. The interactions between each components will be tested before next step.
2. Apply Agent data
After building the basic model which includes the correct logic, the agent behavior needs to be defined to a more detailed level. The results from Phase 1 will be applied to the model to define some variables, such as the distributions of the amount of cars arrived at the carpark. This step will make the agents’ behavior more realistic and the model more accurate.
3. Verify Model
The building of the model is almost done after step a) and b). We will use the model to run certain settings which is very close to one historical data set, the final results will be compared to the actual data set to help us understand the accuracy of the model. Some adjustments may need to be done if the model’s result varies too much from the actual data. The adjustments can be made on global variables, agent behavior or constants.
4. Apply Model
The model is officially done after step c). The user can use the model to verify current design of sites and carparks or the planning of any future developments. The model should be able to illustrate the usage of carpark, whether there are enough slots or otherwise how much the overspill is. After adjusting the carpark size, the user can further find out what is the optimal carpark size for a specific parking which will be very helpful in making plans of future sites development.
Tools
NetLogo is a programmable modeling environment for simulating natural and social phenomena.
NetLogo is particularly well suited for modeling complex systems developing over time. Modelers can give instructions to hundreds or thousands of "agents" all operating independently. This makes it possible to explore the connection between the micro-level behavior of individuals and the macro-level patterns that emerge from their interaction.
Limitation and Assumptions
- Though in total of 65 sites data are collected for analyzing, the amount of data collected and quality of those data for each site is quite deficient. For instance, for each site, only two days data is collected, of which one data set is collected for weekday scenario, the other data set is for weekend scenario. In other words, only one data set is available for analyzing a site in a peak or non-peak day. This deficiency in data quantity may affect the accuracy of results provided by our simulation programme.
- Data is collected every 15 minutes interval within a day instead of continuously for a period of time as this might affect the accuracy of the model and analysis.
