University of Nebraska - Lincoln

DigitalCommons@University of Nebraska - Lincoln

Computer Science and Engineering: Theses,
Computer Science and Engineering, Department of
Dissertations, and Student Research

7-2011

A Reservation-based Smart Parking System

Hongwei Wang
University of Nebraska-Lincoln, [email protected]

Follow this and additional works at: http://digitalcommons.unl.edu/computerscidiss

Part of the Computer Engineering Commons, and the OS and Networks Commons

Wang, Hongwei, "A Reservation-based Smart Parking System" (2011). Computer Science and Engineering: Theses, Dissertations, and
Student Research. 29.
http://digitalcommons.unl.edu/computerscidiss/29

This Article is brought to you for free and open access by the Computer Science and Engineering, Department of at DigitalCommons@University of
Nebraska - Lincoln. It has been accepted for inclusion in Computer Science and Engineering: Theses, Dissertations, and Student Research by an
authorized administrator of DigitalCommons@University of Nebraska - Lincoln.
A RESERVATION-BASED SMART PARKING SYSTEM

by

Hongwei Wang

A THESIS

Presented to the Faculty of

The Graduate College at the University of Nebraska

In Partial Fulfilment of Requirements

For the Degree of Master of Science

Major: Computer Science

Under the Supervision of Lisong Xu and Wenbo He

Lincoln, Nebraska

July, 2011
 A RESERVATION-BASED SMART PARKING SYSTEM

Hongwei Wang, M. S.
University of Nebraska, 2011

Adviser: Lisong Xu and Wenbo He

In metropolitan areas, parking management influences drivers search time and cost
for parking spaces, parking revenue, and traffic congestion. The wide deployment
of wireless parking meters with sensing and communications capabilities allows
the parking authority to monitor the state of each parking space in real time and
optimize the parking management.
In this thesis, we study state-of-the-art parking policies in smart parking
systems, and show that the smart parking system needs to be “smarter”. Our
design goals of the smart parking systems include: (1) simplify the operations of
parking systems, (2) improve drivers’ satisfaction, (3) increase parking revenue,
and (4) alleviate traffic congestion. Through analysis and simulations, we first
show that the proposed reservation-based parking policy has the potential to
achieve the above goals. We then model the behavior of both service providers
and drivers in smart parking systems, and explore the dynamic pricing scheme to
achieve the goals in smart parking system design.
Furthermore, we design and implement a prototype of Reservation-based Smart
Parking System (RSPS) that allows drivers to effectively find and reserve the vacant
parking spaces. With the real time tracking of parking status via various sensing
technologies, a smart parking system will dynamically update the parking price
according to the physical parking status, and the parking price will affect drivers
decision on parking slot selection, therefore, affect the parking status. A smart
parking system can be regarded as a full-fledged cyber-physical system (CPS).
Through extensive experiment based on real traffic traces and a real-world parking
map, the results show that the proposed reservation-based parking policy has the
potential to simplify the operations of parking systems, as well as alleviate traffic
congestion caused by searching for parking.
 iv

ACKNOWLEDGMENTS

I am heartily thankful to my supervisor Dr. Wenbo He for accepting me as her

student and for providing me the opportunity to work on the cutting-edge project
of Cyber-Physical Systems. I would also like to thank her for her encouragement,
guidance and unfailing support throughout the completion of my research project
and this thesis.
I owe my deepest gratitude to my co-supervisor Dr. Lisong Xu for taking
time out of his busy schedule and providing his valuable expertise and guiding
me through the completion of this thesis. I would also like to thank Dr. Byrav
Ramamurthy for agreeing to be on my committee and for providing valuable
feedback and comments on my thesis.
I am grateful for my dearest friend Fan Zhang for his motivation and encour-
agement in a number of ways. I would also like to thank Landu Jiang for always
being there for me
I would like to thank my parents for always being supportive of me and
believing that I could get through this.
Lastly, I offer my regards and blessings to all of those who supported me in
any respect during the completion of this project.
 v

Contents

Contents v

List of Figures viii

1 Introduction 1
1.1 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 A Reservation-based Solution . . . . . . . . . . . . . . . . . . . 3
1.1.2 Dynamic Pricing Scheme . . . . . . . . . . . . . . . . . . . . . 4
1.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Background and Related Work 6

2.1 Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 State-of-the-art Parking Management Solutions . . . . . . . . . . . . . 7
2.3 Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.1 Cyber-Physical System (CPS) . . . . . . . . . . . . . . . . . . . 10
2.4.2 Bay Area Rapid Transit . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.3 Parking Lot Detection . . . . . . . . . . . . . . . . . . . . . . . 13
2.4.4 VANET-Based Smart Parking . . . . . . . . . . . . . . . . . . . 14
 vi

2.4.5 Resource Allocation . . . . . . . . . . . . . . . . . . . . . . . . 15

3 Reservation-based Smart Parking Management Solution 16

3.1 System Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Model of User Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Dynamic Pricing Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3.1 Usage Price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3.2 Statistical Price . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3.3 Congestion Price . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4 System Design 26
4.1 Hardware Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Software Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

5 Evaluation 34
5.1 Simulator Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.1.1 Real Map Import . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.1.2 Parking Demand Modeling . . . . . . . . . . . . . . . . . . . . 35
5.1.3 Simulator Implementation . . . . . . . . . . . . . . . . . . . . . 36
5.2 Simulation Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.3.1 Traffic Searching for Parking . . . . . . . . . . . . . . . . . . . 39
5.3.2 Revenue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.3 Service Differentiation . . . . . . . . . . . . . . . . . . . . . . . 44

6 Conclusion and Future Work 46

6.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
 vii

6.2.1 Synchronization . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2.2 System Deployment . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2.3 Security and Privacy . . . . . . . . . . . . . . . . . . . . . . . . 47

Bibliography 49
 viii

List of Figures

3.1 Integrated Smart Parking Management System . . . . . . . . . . . . . . . 17

4.1 Data Communication among Components . . . . . . . . . . . . . . . . . 28

4.2 Architecture of Parking Management System . . . . . . . . . . . . . . . . 30
4.3 Architecture of Parking Management System . . . . . . . . . . . . . . . . 31
4.4 Architecture of Parking Management System . . . . . . . . . . . . . . . . 32

5.1 Map of Los Angeles Downtown . . . . . . . . . . . . . . . . . . . . . . . . 37

5.2 Incoming and Outgoing Traffic Flows . . . . . . . . . . . . . . . . . . . . 38
5.3 Traffic Searching for Parking Comparison under Different Parking Guid-
ance Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.4 Traffic Searching for Parking under Different Pricing Schemes . . . . . . 40
5.5 Revenue Comparison between Dynamic and Fixed Pricing . . . . . . . . 42
5.6 Revenue Comparison between Dynamic and Random Selection of Fixed
Pricing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.7 Failure Rates of Users with Different Budgets . . . . . . . . . . . . . . . . 44
 1

Chapter 1

Introduction

The study of Cyber-Physical System (CPS) has become a key area of research.
It refers to a new generation of systems with integrated computation and com-
munication capabilities that allow users to interact with the physical world. The
ability of interaction with physical world is a key catalyst for future technology
development. CPS, therefore, provides significant opportunities for design and
development of next-generation traffic management solutions. As an important
component of traffic system, parking management system is playing an important
role and affecting people’s daily life. By detecting and processing the information
from parking lots, smart parking system allows drivers to obtain real-time parking
information and alleviates parking contentions, which is a practical application of
CPS.
The parking industry generates billions of dollars in annual revenue in the
United States alone, and parking regulations may affect people’s concerns about
traffic congestion, air pollution, drivers’ frustration about parking searching, and
municipal objectives. For instance, a recent survey [29] shows that during rush
hour in most big cities, the traffic generated by cars searching for parking spaces
 2

takes up to 40% of the total traffic. Therefore, in these densely populated urban
areas, a certain amount of traffic congestion and delay are due to parking. A recent
study [25] shows, in a business district of Los Angeles, vehicles looking for parking
burn 47,000 gallons of gasoline and produced 730 tons of carbon dioxide, which
is equivalent of 38 trips around the world. Clearly, the problems associated with
parking imposes significant societal costs, both economically and ecologically [18].
In order to address these problems associated with parking, smart parking
systems aiming to satisfy the involved parties (e.g., parking service providers and
drivers) have been developed. However, most current smart parking or parking
guidance systems [1, 2] only collect and publish live parking information to direct
drivers to available parking spaces near their destinations. These systems are
not “smart” enough, because they cannot successfully help drivers find a desired
parking space in crowded areas, and sometimes make the situation worse. For
example, if available spaces in a congested area are less than the spaces in demand,
more drivers trying to park will head for the limited available spaces, causing
severer congestion. In this case, detailed information associated with parking
availability would allow drivers to make better decisions on use of parking lots
and road-side parking.
In contrast to such parking information guidance systems, this thesis presents
a Reservation-based Smart Parking System (RSPS) that not only to broadcast real-time
parking price based on the parking availability to the drivers as part of a communal
application, but also to provide reservation service as part of user-targeted service.
Built on advanced sensing and mobile communication techniques, RSPS processes
streams of timestamped sensing data from sensor network in parking lot, calculates
the real-time parking price based on parking availability information and publishes
the parking price to the drivers. On the other side, the drivers can retrieve parking
 3

price and reserve their desired vacant spaces via Wi-Fi or Internet.

1.1 Approach

As the technical foundation of smart parking system, computing devices (e.g.,

smart phones, wireless sensors and personal laptops) turn progressively smaller,
cheaper and more powerful. As a result, mobile and pervasive computing is
becoming an indispensable component in the distributed networked computing in-
frastructures. It provides us a powerful platform to compute real-time information
from physical world (physical part), as well as communicate with people (cyber
part). Cyber-physical systems are so involved in the everyday life that we should
not underestimate their impacts. In this thesis, the objective of smart parking
system is to develop technologies which will help to build a mobile and real-time
computing system which make the computing and communication capabilities
fully play, and elegantly integrated with users.

1.1.1 A Reservation-based Solution

Based on the observation that parking space reservation can help drivers to reduce
the search time dramatically, we propose a reservation-based solution, built on
advanced sensing and mobile communication technology, with the objectives to
alleviate the parking contention, balance the benefits between parking service
providers and drivers, coordinate among service providers, differentiate the needs
of individual drivers, and reduce the amount of traffic searching for parking as
well. To achieve the design goals, a powerful tool to model the behavior of both
service providers and drivers is required. Meanwhile, we need to design control
signals to guide the parking selection of large scale, autonomous drivers.
 4

1.1.2 Dynamic Pricing Scheme

Normally, a parking system manages the parking resources, and the drivers buy
or reserve the permits to park somewhere they want within a specific parking
district for the chunk of time they purchase. In the proposed model, the parking
management system adopts dynamic pricing scheme to generate prices for parking
spaces in different parking lots. The parking price reflects the real time parking
availability. It not only serves as a control signal to balance the parking lot
utilization, but also improves the revenue for service providers. On the other
hand, drivers rely on the utility functions to determine where to park (e.g., select
the most convenient parking spaces) under a certain budget constraint. Through
the simulations, we will show that: the amount of traffic searching for parking
is reduced, the solution guarantees the service quality for different users and the
revenue is increased for service providers.

1.2 Contributions

In this thesis, we propose a reservation-base smart parking management solution.

It implements a dynamic parking price scheme and reservation parking service.
Moreover, we design and implement the prototype of a Reservation-based Smart
Parking System (RSPS) to evaluate the proposed reservation-base smart parking
management solution. We summarize the contributions made by this work as
follows.

• The RSPS provides parking reservation service, which focuses on the entire
parking condition in certain area, instead of single parking lot. With the
real-time reservation service, the drivers can find and reserve their desired
 5

vacant parking spaces quickly. Therefore, the gasoline and time in search of
vacant parking space is reduced.

• We design the dynamic parking price scheme for parking service. In order
to satisfy the different concerns from different parts in parking system, the
dynamic price scheme is designed to generate more parking revenue, provide
differentiated services for different drivers and reduce the parking congestion.
In this thesis, we demonstrate that it is able to reduce the traffic congestion
caused by parking searching.

• We demonstrate the usefulness and effectiveness of RSPS by implementing a

prototype. In this project, we design and develop the three-tier architecture,
(1) data tier: deployed sensor network to collect real-time parking informa-
tion, (2) logic tier: dynamic pricing engineer in process of sensing data, (3)
representation tier: web service.

1.3 Thesis Organization

The rest of this thesis is organized as follows. In Chapter 2, we set the background
of the proposed research by introducing performance metrics and several exist-
ing approaches for smart parking systems, and summarize the related work. In
Chapter 3, we present the details of Reservation-based Smart Parking (RSP) man-
agement solution. In Chapter 4, we present the detailed architecture of proposed
reservation-based smart parking system. In Chapter 5, we evaluate the proposed
RSP management solution through extensive simulation. Finally, we conclude this
thesis and discuss the future work in Chapter 6.
 6

Chapter 2

Background and Related Work

In this thesis, we mainly focus on a management system that assists drivers to find
parking spaces in a specific parking district, and satisfies the needs of both parking
providers and drivers. In addition, an important goal of the system is to reduce the
traffic searching for parking, hence reduce energy consumption and air pollution.
In this chapter, we review background on smart parking systems, including the
performance metrics, existing solutions and challenges. We also briefly discuss the
related work.

2.1 Performance Metrics

To evaluate the performance of a parking management system, we introduce the

following metrics, which reflect the needs of involved parties, and our concerns on
traffic congestion and environmental protection.

• Walking Distance: Walking distance is defined as the distance from a driver’s

selected parking space to the destination. It is an important factor for a driver
to determine where to park. Usually, a driver wants to park as close to the
 7

destination as possible if his budget permits. Therefore, the walking distance

indicates the satisfaction of drivers.

• Parking Revenue: Regardless of whether a parking lot is privately owned or

municipally owned, parking revenue represents the benefit to the parking
providers. Since multiple parties are involved in this system, our design
does not aim to maximize the parking revenue for service providers only,
but allows them to obtain profits at reasonable level.

• Service Differentiation: Usually a driver wishes to pay as little as possible, and

has a certain budget for parking. In a crowded area, the parking resource
is limited. To alleviate the contention on parking resource and maintain
reasonable parking revenue for service providers, the management system
should differentiate the drivers according to their budget and need.

• Traffic Searching for Parking: The traffic generated by drivers searching for
parking is not negligible and reflects the social welfare. Hence, an efficient
parking guidance system should efficiently reduce the traffic searching for
parking. Also, reducing the amount of searching time for parking is desired
by drivers.

We investigate performance of the proposed smart parking system using these

performance metrics.

2.2 State-of-the-art Parking Management Solutions

We now introduce several existing parking guidance approaches and show their
limitations. We simulate the parking system performance under different parking
management strategies and show results in Chapter 5.
 8

• Blind Search: Blind searching is adopted by users when no parking informa-

tion is available. So drivers search parking spaces randomly within a certain
distance to their destination. If a driver gets an available space, the driver
will stop searching; otherwise, the driver will continuously searching in the
neighboring parking lots until he finds space. In this case, there is no control
signal to guide drivers’ behavior.

• Parking Information Sharing (PIS): This mechanism represents the current state
of the smart parking system design. When a driver obtains the parking
availability information near his or her destination, the driver will know if
the desired parking lot has available spaces. Hence, individual drivers make
the decision according to the parking availability information. If a parking
lot has a very few parking spaces available in busy hours, it is likely that
more drivers struggle for less parking spaces. This phenomenon is called
“multiple-car-chasing-single-space”, which may cause severe congestion.

• Buffed PIS (BPIS) To alleviate the “multiple-car-chase-single-slot” phenomenon,

some designers of smart parking systems have devised a solution to leave
a buffer when publishing the live availability information. Therefore, if a
parking lot has less unoccupied spaces than a threshold, the system will
show that the parking lot is fully occupied. But it is difficult to determine the
threshold for the buffer. If the buffer is too small, the problem of “multiple-
car-chase-single-space” will not be eliminated. If it is too large, the utilization
of parking spaces will be low.
 9

2.3 Challenges

Given the design objectives of smart parking systems that requires the coordination
among multiple parties, we summarize the main design considerations as follows:

• Performance v.s. Overhead: In order to complete parking space reservation,

drivers have to communicate with the system. However, there is a tradeoff
between the overhead to generate and convey control signals to drivers v.s.
performance optimality. For example, if the system computes the fine-grained
control signals for individual drivers and adopts unicast to inform the drivers
about parking system status and instructions, the overhead will be very large.
Hence, the parking management system generates a uniform control signal
(e.g., the utilization of parking lot and parking price), and broadcasts the
control signal to drivers.

• Trade-off Between Benefits to Drivers and Service Providers: Multiple parties

(drivers and service providers) are involved in the parking system operation.
The state of the system depends on their interaction with each other. To
balance the needs of involved parties, we use parking price as the control
signal to coordinate the involved parties.

• Differentiated Service for Large Scale Autonomous Drivers: Thousands of drivers

make parking decision autonomously. They have different needs and bud-
gets for parking and their interpretation of parking information is different.
Providing differentiated service for drivers is important to satisfy individual
users. In this sense, the service quality is determined by parking price.

As alluded to above, the blind search system is an open loop system, where
users make decision without looking at the state of the system. The PIS and
 10

BPIS strategies allow drivers to make decisions based on the system state (e.g.,
parking availability information). However, since the drivers can obtain the infor-
mation of parking availability, they may pursue limited parking spaces in busy
district resulting more traffic contention. To solve this phenomena of multiple-
car-chase-single-space and reduce the traffic searching for parking, we suggest a
reservation-based system, where drivers make reservations through the parking
management system. If a driver makes the reservation successfully, it guarantees
an available parking space for him, and the driver can park at the reserved space
without searching. The reservation-based system allows drivers to select the most
convenient parking space under their budget constraints. To address the design
challenges, we use a dynamic pricing scheme to determine the reservation charge
for parking. This will regulate the parking behavior of drivers, and fulfill the
needs of service providers. The detailed model for drivers and service providers is
introduced in Chapter 3.

2.4 Related Work

Currently, most research work on smart parking is from the perspective of system
design, which focus on implementing a wireless sensor network to detect parking
information and provide real-time parking service. In addition, we introduce the
pricing-related topics in networks, which provide us a powerful tool to manage
parking lots.

2.4.1 Cyber-Physical System (CPS)

As the frontier of current research, CPS is revealing numerous opportunities and

challenges, and encouraging multidisciplinary collaborative research between
 11

academia and industry [30]. Although it is difficult to make a clear definition of

CPS, it is consider as a physical and engineered system integrated with monitored,
coordinated and controlled operations, as well as a computing and communication
core[21]. From the nano-level to the world-scale systems, the cyber-world of
computing and communications with the physical world is transforming the way
that how humans interact with and control the physical world around us.
The innovative applications, such as [12], in CPS have great social and econom-
ical impact for national competitiveness. Many researchers focus different areas of
CPS, including transportation system [23], biomedical and healthcare systems[13],
defense and aerospace industry, and smart grid and renewable energy [6]. As
the major energy consumer, the revolution of transportation holds great promise.
According a recent survey in 2010[4], the transportation consumption of total
energy in U.S. reached 29.2% in 2008, which is the highest record since 1973. And
88% of these trips are by car in U.S. [3]. As a result, daily commuting represents a
significant part of total transportation consumption. The research on transportation
of CPS focuses on the smart transportation systems, which is aiming to reduce
energy consumption, and driverless systems. In [15], the author investigates the
optimal transportation emission reduction strategies and identifies optimal ways
to reduce transport energy consumption and pollution emissions. Furthermore,
the DARPA Grand Challenge [12], which is a competition for driverless vehicles,
is an important application of CPS. The researchers from different institutions
design and develop a variety of different software and hardware combinations for
interpreting sensor data, planning, and execution.
Since CPS can be considered as a confluence of distributed embedded systems
and real-time system [21] interacting with physical world, the precision of com-
puting and communication is interfaced with the uncertainty and noise in the
 12

physical environment. There are several research challenges for CPS, including
(1) CPS composition, (2) robustness, safety, and security of CPS, (3) control and
hybrid systems, (4) computational abstractions, (5) architecture, (6) real-time em-
bedded systems abstractions, (7) sensor and mobile networks, (8) model-based
development of CPS, (9) verification, validation, and certification of CPS and (10)
education and training.
As an important application of CPS, transportation cyber-physical system [5, 32,
31, 9] enables people to reach their destination safely and reliably. In [5], the author
mainly discusses the cyber-physical aerospace, which is relevant to the challenges
and future directions in transportation and exploration systems. The transportation
cyber-physical system improves the dramatic safety and transportation efficiency,
such as anti-lock brakes and advanced flight management systems. The air
transportation network is an example of transportation cyber-physical system.
The air traffic is suffering adverse weather and density problem. The CPS should
provide an adaptive flight planning capability with seamless integration of strategic
and tactical reasoning system. In [32], the author describes the impact of the
mobile internet to change the transportation cyber-physical system. With the rapid
development of mobile technology in last five years, it is constructing the dedicated
infrastructure systems to monitor traffic. The Smartphone with localization module
is changing the traffic monitoring system. It is able to provide the real-time traffic
information for entire transportation network. As one of the first instantiation
of transportation cyber-physical systems, the traffic information system is being
progressively influenced by mobile technology. On the other hand, there are
still some promising and fundamental challenges of mobile device technology
remaining to be solved. In [31], the author presents the limitations of current
automotive Cyber-Physical System. Because of its vehicle-centric view, it only
 13

provides partial information about surrounding environment. Therefore, it is not

fit for satisfying the needs of embedded human in the system. The new automotive
CPS should be more open and flexible to develop new human mobility models.
In [9], the author discusses substantial technological and engineering advances of
transportation CPS, it provides the advanced autonomous control choices for cars,
trains, and aircraft. With the correct automative control, it ensures a safe operation
of the control assistants under all circumstances.
Another instantiation of CPS is advanced electric power grid [19, 21], which is
one of the largest complex interconnected networks. In [19], the author introduce
the new power grid, power electronics, and embedded control move, based on
concepts of CPS, beyond centralized monitoring to active, distributed, control of the
power grid. In [21], the author presents some new challenges of developing a fully
integrated robust, stable, failure-free advanced power network with interacting
distributed, intelligent real-time control through the composition of cyber and
physical resources. The specific aspects include multiple time scales of interacting,
distributed and control, high resilience, security policy and sustainable power
system.

2.4.2 Bay Area Rapid Transit

As an important application of CPS, smart parking management systems have been

designed to more efficiently use parking capacity. The smart parking management
field test at the BART station in the East San Francisco Bay Area is good example to
demonstrate the feasibility of smart parking systems and describe the specification
of its technology [24]. The authors complete a comparative evaluation of rider
attributes at three BART stations to be identified as the potential sites. The data
 14

from drivers at three stations provide the insight of potential demand for smart
parking system. Since the drivers also appear to use the Internet frequently, it
demonstrates the potential of smart parking reservation service, which will be
detailedly discussed in Chapter 3. In order to evaluate the smart parking field,
they implement two user interfaces: two VMSs that display parking availability
information, and a centralized intelligent reservation system. The survey results
from their test suggest that a potential market for a daily paid parking service.
However, only providing availability information may cause some other serious
problems, such as the “multiple-car-chase-single-slot” phenomenon.

2.4.3 Parking Lot Detection

As a part of Smart Parking System, the parking lot detection has been developed
by Jake Reisdorff et al. [22] as a course project under the supervision of Dr. Sharad
Seth in Computer Science & Engineering at University of Nebraska-Lincoln. The
project is aimed to providing a parking garage detection solution that is accessible
by a webpage. Specifically, they focus on the Stadium Drive parking garage on the
UNL city campus on which to test their project. They use a mounted camera in the
garage to take images from a static position, and send to a web server. The server
runs an image processing application to detect if the parking spaces are available
or not, and deliver the raw data of image process to the web application. Based on
the raw data, the web application determines the number of unoccupied spaces,
and show a map of parking garage with availability information.
 15

2.4.4 VANET-Based Smart Parking

With development of Vehicular Ad-Hoc Network (VANET), VANET-Based pro-

vides another option for smart parking systems. In [8], the authors present a smart
parking management system based on wireless sensor network technology, which
provides remote parking monitoring, automated guidance and parking reservation
service. They demonstrate this system architecture can help commuters to find
vacant parking spaces. Rongxing Lu et al. [16] introduce a new Smart PARKing
(SPARK) scheme, which is based on Vehicle Ad Hoc Networks (VANET), provides
drivers with accurate and convenient parking services in large parking lots, in-
cluding real-time parking navigation, intelligent antitheft protection, and friendly
parking information dissemination. In [7], authors proposed a scalable information
dissemination algorithm for discovery of free parking spaces via VANET.

2.4.5 Resource Allocation

Furthermore, appropriate parking control to motivate drivers to balance their

parking demands is another goal for smart parking system. Pricing policy, as
an important economic tool, is effective to allocate resource in network. In [11],
Kelly et al. show that, when selfish users seek to maximize their benefits, the
system can provide incentives to reach the optimization of global network. In
[28], authors proposed a reasonably complete DiffServ pricing model integrated a
pricing structure for different service class and the demand behavior of users. They
show that, when different service classes operate at different target utilization,
they provide different levels of service, and users see a stable service price and
maintain stable. Moreover, Feldman et al in [10] formulate the fixed budget
resource allocation game and study the existence and performance of a distributed
 16

market-based resource allocation system.

 17

Chapter 3

Reservation-based Smart Parking

Management Solution

In this chapter, we present the proposed Reservation-based Smart Parking (RSP)

management solution, which implements a reservation policy and dynamic pricing
scheme.

3.1 System Architecture

Fig. 3.1 shows three components in the system, including parking lots, users
and the management system. The management system determines the dynamic
parking prices based on real-time parking information, and broadcast live parking
prices to users (also drivers). The price reflects the relationship between demand
and supply, and implies the congestion level. Upon receiving dynamic parking
prices, the user selects a desired parking lot and reserves a space in the parking lot.
According to his budget constraint and convenience degree, the parking decision
would vary by user. As a result, the state of parking resources is changed by users’
 18

Users Management
System

Price Database
Parking Prices

Price Decision

Infornmation Analysis

Parking Lots

Parking Resources

Reservation Sensor Network

Sensing & Reservation
Requests information
Resevation Authority

Figure 3.1: Integrated Smart Parking Management System

parking decisions. The parking lot consists of a group of parking spaces. The
on-street parking can also be considered as a virtual parking lot. The state of a
parking lot is the number of occupied spaces versus total spaces. Every parking
lot has access to the Internet to communicate with the management system and
users, and share parking information with other parking lots. In each parking
lot, the reservation authority is deployed for authenticating the individual user’s
identity and reservation request. In this case, the reservation authority in the
parking lot communicates with the specific user individually. Once the reservation
order is confirmed, the reservation authority updates reservation information to
hold the related space for the user. The sensor system deployed in parking lot is
responsible for monitoring the real-time condition of parking lots and delivers the
live aggregated sensing information (the number of available spaces or occupancy
rate) to the management system. The sensing information is updated on demand.
 19

Upon retrieving the parking information, the management system updates the
state of the parking lot. Based on the state of parking lots, the management
system (1) analyzes their occupancy status and congestion level, (2) determines the
parking prices according to dynamic pricing scheme, (3) broadcasts the prices to
all users periodically, and (4) stores the parking information and prices for further
analysis. The management system serves as the centralized decision-making body
in a planned economy. It makes all pricing decisions regarding the state of parking
lots and user demands [20]. This system is a closed-loop system to dynamically
adjust parking price, balance the benefits between users, and service providers
and reduce traffic searching for parking.
By placing the reservation authority in individual parking lots, we simplify a
lot of issues related to the implementation, including communication overhead,
reservation synchronization and load balancing. Since each user only has to
communicate with his desired parking lot to make his reservation, rather than
the centralized management system, the communication overhead of reservation
is highly reduced. Also, since each parking lot manages its own reservation
information, it makes the reservation requests from users easily to be synchronized,
comparing with reservation synchronization in the management system.

3.2 Model of User Behavior

To model user behavior, we use the framework of the competitive market model
[27], which includes two agents: consumers and producers. To describe the
relationship between service providers (also producers) and users (also consumers),
the price of parking service is defined as the exchange rate of parking spaces,
dynamically determined by parking demands and congestion level. Given dynamic
 20

parking price, a user selects a parking space under his budget constraint. Upon
receiving the prices of parking lots that have available spaces, the user selects
a convenient lot to park from those whose prices are acceptable. In order to
model the parking decision behavior, we define the utility function to balance the
convenience and cost in the parking process.
In the real world, users desire to maximize the convenience and minimize the
parking cost, though trade-off must be made. Here, given the parking price p of

specific parking lot, the relative cost is defined as a function of p, C ( p) = κ p b,
where κ denotes the user’s sensitivity to the price and b is the budget constraint
of the user. In this case, a user with higher budget will have more tolerance for
higher price and tend to pursue the parking space providing more convenience.
Meanwhile, the convenience can be measured by walking distance d, hence the
gain of a user is a function of d,

dm
B (d) = υ log , for d ≤ dm (3.1)
d

where B represents the convenience if the user selects a parking lot with the
walking distance d, dm denotes the maximum walking distance the user can
tolerate, and υ represents the sensitivity to walking distance. As represented in the
literature [11][28], when all user utilities are defined as logarithmic, the optimal
solution becomes proportionally fair. Accordingly, the optimization of the utility
function for a user’s task of parking selection in an area that contains multiple
 21

parking lots, is represented as

h    i
max U = max B d j − C p j
j j

pj
 
dm (3.2)
= max υ log j − κ
j d b

s.t. p j ≤ b d j ≤ dm

where p j represents the parking price at jth parking lot and d j is the distance from
jth parking lot to the user’s destination. Note that, since users make their parking
decision during one time slot, the utility function is only for a given time slot.
For an individual user, the parking decision is to maximize the utility, which is a
function of parking price and walking distance. The user’s decision cannot have
impact on other user’s decision, although the parking price may be modified by the
change in demand of the users. Furthermore, in reality, a user with high budget
may desire more convenience, which means the maximum walking distance he
can tolerate may be less. In our model, to avoid computation complexity, we
use the constant values for maximum walking distance and budget to imply this
relationship.

3.3 Dynamic Pricing Scheme

Service providers that own the parking resources (parking lots) collect real-time
parking information to determine the parking prices, and broadcast them to the
drivers to help them make reservations. From the perspective of service providers,
dynamic pricing scheme is for generating more revenue from the parking service,
supporting service differentiation for users with different requirements and bud-
gets, and reducing the traffic searching for parking. On the other hand, users desire
 22

to learn parking information from the parking price and obtain the guarantee of
service quality.
In order to achieve these design goals, the formulation of dynamic pricing is
decomposed into usage price (pu ), statistical price (ps ), and congestion price (pc ).
Specifically, the usage price is determined by the real-time state of parking lot
aiming for reflecting real-time parking condition and increasing the revenue; the
statistical price is measured by the historical price, which provides users with a
reasonable price for their reservation time period and reduce the potential traffic
of parking searching; and the congestion price reflects the congestion in near future,
and prevents the sudden increase of the parking demand. Both usage price and
congestion price are computed periodically, with the interval τ, statistical price is
computed on demand. The parking price p is represented as

p = β u pu + β s ps + β c pc (3.3)

where β u , β s , β c represent the different weights for different price components.

Note that the combined parking price p is the average price for total reservation
time.

3.3.1 Usage Price

The usage price is determined by resources consumed, specifically the occupied

parking spaces, and users’ potential demands. The usage price ( pu ) is set such that
it can maximize the revenue, corresponding to various costs of service provider
associated with real-time state of resource. Given an individual parking lot with
J parking spaces, the usage price during ith time slots is piu , which is limited by
number of users demanding parking. The pricing decision is to obtain the optimal
 23

profit: " #
I
max
piu
∑ (xi piu − f ( ji , J ))
i (3.4)
subject to: xi , ji ≤ J

where xi represents the total demand for parking spaces with price piu during
ith time slots, I is the total time slots, and f ( ji , J ) is the function of parking cost
associated with occupied parking spaces ji and total parking spaces J.
According to the utility function of users described above, users select their parking
lots based on walking distance and parking price. However, since users estimate
walking distance depending on their own capability and other unexpected factors,
the service provider cannot learn this behavior from the users. So we simply
assume that the total demand xi for parking spaces with price piu is determined by
the price, which can be represented as

Ci
xi = (3.5)
δi piu

where Ci represents the total number of users whose budgets are higher than piu
and δi is the parameter of preference to choose a parking lot with price piu .
Denote the basic price of parking service with the lowest utilization by pbasic . ϕi
is used for adjusting the basic price based on real-time state of parking lot, the
service price can be estimated as piu = ϕi pbasic /(1 − σi ), where σi is the utilization
of specific parking lot. The effective demand associated with σi is Ci /δi piu (1 − σi ).
Accordingly, the Equation (3.4) can be written as

" #
I
Ci i ϕi pbasic
  
max
piu
∑ p
δi piu u
− f j i
, J , p i
u =
(1 − σ i )
i (3.6)
Ci
s.t. i i ≤ J, ϕi ≤ 1
δ p u (1 − σ i )
 24

The optimal problem of the Lagrangian is represented as

" #
I
Ci Ci
    
max
piu
∑ δi
+λ J− i i
δ ϕ pbasic
− f ji , J (3.7)
i

The optimal solution is

Ci
piu = (3.8)
δ i J (1 − σ i )

3.3.2 Statistical Price

The statistical price is set to predict the parking price beyond the current time
slot. If a user reserves a parking space during the upcoming time slots, the
service provider will impose an opportunity cost by depriving other users of the
opportunity to be admitted to use that parking space. However, the live usage
pricing cannot correctly reflect the opportunity cost in the future. In this case, the
service provider has to provide a new fair charge to predict future parking price,
instead of selling resources at the current usage price. Since the total traffic flows
are periodic, the parking conditions are similarly changing. Therefore, the historic
data provides the service provider a powerful tool to predict the parking price
over future time slots. So the statistical price ( pis ) of parking reservation service is
defined as
I0 h i avg i
i
( k −1)
∑ $pu + (1 − $) pu
i
pis = (3.9)
I0

where pis is the statistical price for the ith time slot during the current time period
kth (we use one day works as one time period that consists of I time slots), I 0 is the
i avg
total time slot that user reserves, pu is the average usage price at the ith time slot
i
in all last periods, pu(k−1) is the usage price at ith time slot during last (k − 1)th time
period, and $ is weight parameter.
 25

By providing users this statistical price, the service provider lets users learn the
future parking price, which can prevent potential congestion. At the same time,
since the statistical price can predict the future usage price, this more reasonable
price allows the service providers to increase the revenue and users’ satisfaction.

3.3.3 Congestion Price

To avoid many users pursuing a few parking spaces and alleviate traffic congestion
caused by parking searching, we propose an additional price component, which
is congestion-sensitive to encourage users to balance their parking selections and
reduce congestion. The congestions considered are traffic congestion and parking
congestion. Since the traffic congestion cannot be monitored by parking sensors
directly, and many traffic congestions are not caused by parking searching, we
only focus on the growth rate of parking occupancy to indicate traffic condition.
If the traffic searching for parking spaces gives contribution to traffic congestion,
the rate of parking occupancy has to grow dramatically in short time. Moreover,
when the available parking spaces are lower than a certain proportion, it will cause
parking congestion and conflicts if too many users pursue these spaces. Therefore,
two kinds of congestion pricing are considered: pricing when the growth rate of
occupancy exceeds a certain threshold, and pricing when the parking occupancy
reaches certain level. Both pricing are calculated iteratively to prevent the price
highly oscillating. This kind of change in price, increasing when the supply is
below users’ demand and decreasing when the supply exceeds users’ demand,
can regulate the relationship of supply and demand to reach certain equilibrium.
In this paper, we classify the congestion level with congestion factor l. When the
congestion factor exceeds the threshold of certain level, an additional congestion
 26

price is charged. The congestion price pic during ith time slot is given by

   +

( i −1) dr ri − riavg 
pic = pc + ξl + ωl ,0 (3.10)
 di riavg 

where ξ l and ω l are the congestion parameters to adjust the weight according the
different congestion factors l, and ri and riavg represent the occupancy rate and
average occupancy rate during ith time slot.
During the congestion, the users have to suffer the extra congestion charge, or
select other parking lots with lower parking charge. So Equation (3.10) drives
the user demand towards the parking lots with lower occupancy rate and less
growth rate. The provider applies the congestion charge when the congestion
reaches certain level, and adjusts the congestion charge and level dynamically
according to the real-time parking information. After the congestion is removed,
the congestion price is decreased gradually to zero to avoid further congestion and
price oscillation if new congestion is found.
In this section, we have presented a dynamic pricing scheme consisting of three
different pricing components: usage charge, statistical charge and congestion charge.
Both the usage price and congestion price are adjusted periodically in a given time
scale. And the real-time price information is stored in a database for statistical
price and further analysis. In reality, the users only need to know the total price,
instead of specific price components.
 27

Chapter 4

System Design

With the model details as given above, we now present the design of Reservation-
based Smart Parking System (RSPS). The various components of our design are
closely inter-related. Given this interaction, we first present the design concerns of
the system in this section, before moving on to the details in subsequent sections.

• Reservation Performance: The RSPS utilizes both the Internet and Wi-Fi,
whereby drivers can check the real-time parking information and complete
their reservation. However, there is a bottleneck to the system when many
drivers are simultaneously making reservation. In this case, the system
has to synchronize the parking information and handle each reservation
request, which significantly reduce the system performance, and even cause
some conflicts. In order to address this challenge, we design a distributed
reservation strategy implemented in the proposed smart parking system.
When a driver selects the desired parking lot, the system will reconnect the
driver to the subsystem in related parking lot, the driver can complete the
reservation without communicating with the central system. Therefore, the
central system no longer needs to maintain the reservation service.
 28

• Data Collection and Local Presentation: The system collects and stores the data
about the performance metrics, including the status of parking space, reser-
vation time, parking location, driver’s identity. All data stored by the system
is at least stamped with time metadata. Furthermore, the system allows the
driver to check the parking information, including the location of parking
spaces, the vacancy time of parking spaces and reservation information. In
order to protect the security of the system, we separately design a repository
of sensing data and a mirror database of reservation. The repository is the
sink of the sensing data, and the mirror database is synchronized with the
repository and stores the reservation information. In this way, the drivers are
only able to check and update the information in the mirror database.

• Drivers’ Identity Verification: Once the reserved space is detected to be occu-

pied by a vehicle, the system should verify the driver’s identity. If he/she did
not reserve the space, the system will alarm the driver this space is reserved.
In our proposed system, the drivers can visit the website and verify their
identity via Internet. If they cannot access Internet, the drivers are able to
communicate with the sensors by Bluetooth wireless connection.

4.1 Hardware Design

The system hardware is organized into three main components, the sensor network,
the central server and the mobile device, as shown in Fig 4.1. In the following, we
discuss the detailed design and implementation of each component, along with
the specification of communication between them. In our project, we developed
a number of functions on Zigbee sensors that provide a continuous measure of
parking status for each space. Each sensor is integrated with two wireless mote.
 29

Figure 4.1: Data Communication among Components

The wireless mote platform provides a 250 kpps 802.15.4 wireless radio, 8 channel
A/D and an 8 MHz microcontroller for on board digital signal processing. Mote 1
hosts light and vibration sensors, which is used to detect the vehicles. In reality, the
light sensor is easily interfered by light sources. So we use highly directional beam
to strengthen light and reduce the interference. Mote 2 hosts the communication
module of Bluetooth. As a result, the sensor bridges the communication between
the Zigbee on mote and the Bluetooth module on smartphone (e.g., Android G1).
In this case, the sensor confirm the identity of users when vehicle is detected in
reserved parking lot.
 30

The mobile phone is used to assess Internet, over Wi-Fi or a GSM cellular
network, to obtain the information of parking availability and make parking
reservation from the Internet server. The mobile phone also provides the Bluetooth
module to communicate with sensors when verifying the driver’s identity.
There are a central Ethernet-connected server deployed with storage and compu-
tational power. These servers provide hardware support for the software services,
which are described in section III.C. In particular, it is for system users to request
the services of parking information and parking reservation. Once user’s reser-
vation is authorized, the server will update the state of related parking sensor by
wireless low power link, IEEE 802.15.4.

4.2 Software Design

Fig. 4.2 shows the design of software architecture, primarily defining the iRev,
which is the central location of the system to host applications and functions as the
point of control and configuration for the distributed system. Primary software
elements are discussed in the following.
RSPS has a subsystem of sensor network in the parking lots. The sensor
networks provide the real-time parking information to the upper layer. Here we
categorize sensor nodes to detecting nodes and collecting nodes. Specifically, the
detecting nodes take the responsibility to monitor the status of parking spaces, as
well as communicate with mobile phones. Moreover, the collecting nodes advertise
itself as the root node and are responsible to collect and deliver the detected event
and data from detecting nodes.The sensor nodes are able to achieve end-to-end
connectivity across a set of nodes and access points by implementing collection
tree protocol.
 31

Web Server

iRev

iRev Applications

WiFi Reservation Price Mobile

Monitor Authority Decider Payment

Parking Info
Repository
Mirror Data
Database Sync

Query Stack

Continuous Query Real-time Sensing

Adapter Data Steam

Sensor Network

Figure 4.2: Architecture of Parking Management System

As the middle layer between the sensor network and web server, the iRev is
the sink for all data sent from the lower sensor nodes. To simplify application
development, we have developed query stack for delay-tolerant continuous query
processing. Meanwhile, the parking information repository can retrieve real-time
sensing data stream. To transfer data efficiently and reduce the complexity of
the system, we have developed data synchronization processor that can connect
and synchronize the data between parking information repository and mirror
database. The user can check the parking price and make their reservation via
mirror database, and data synchronization processor takes the responsibility to
transfer related data to repository. In this case, the security of parking information
is protected, as well as the information redundancy is reduced. Based on the
real-time parking data, there are four specific applications in the iRev, including
 32

Wi-Fi monitor, reservation authority, price decider and smart payment. (1) Wi-Fi
monitor is designed to monitor and report on Wi-Fi Access Points. It monitors
and reports on performance, availability and problems. (2) Reservation authority
is responsible for verifying the reservation process and driver’s identity. (3) Price
decider is to determine the parking price based the parking state. (4) Mobile
payment allows drivers to use their mobile phones to pay parking fee.

Figure 4.3: Architecture of Parking Management System

RSPS provides the web service to the drivers, as shown in Fig. 4.3 and 4.4.
When the sensor network in parking lots detects the change of parking states, it
 33

sends the real-time sensing data to the parking info repository. Then, the data
synchronization processor updates the associated data in mirror database based
on parking info repository. The system dynamically updates the parking and
reservation information on the website according to the data stored in a mirror
database. The driver is able to obtain the real-time parking information from the

Figure 4.4: Architecture of Parking Management System

web server, as shown in Fig. 4.3. If the driver selects his/her desired parking lot,
s/he is able to reserve a parking space in the selected parking lot. After verifying
 34

the driver’s identity in reservation authority, the system updates the reservation
information in the mirror database, and the driver receives the confirmation
information, as shown in Fig. 4.4.
 35

Chapter 5

Evaluation

In this chapter, we simulate the proposed Reservation-based Smart Parking (RSP)

system based on real traffic traces and real-world parking map, and demonstrate
the performance of the RSP in terms of the metrics shown in Section 2.

5.1 Simulator Design

In order to investigate the parking guidance policies and the proposed RSP sys-
tem, we have to develop a simulator to import the real-world map and traffic
traces, simulate users’ parking behaviors and implement related parking strategies.
Specifically, we use Java 1.5 to develop the main components, including dynamic
pricing scheme, data access and iRev applications. For the data repository and
minor database, we use MySQL. And we use JSP to implement the web service by
using Apache Tomcat as the web environment. For the sensor program, we use
TinyOS and nesC to develop specific applications.
 36

5.1.1 Real Map Import

This simulator allows us to import a real-world map as the target area, and
acquires the information from the map, e.g., distance and paths. Given the map,
let G = ( N, A) be traffic network defined by a set N of nodes and a set A of edges,
where N and A represent the set of blocks and the set of roads connecting blocks.
With the parking map, we aggregate the parking lots in one block as a virtual
parking lot. Therefore, each node has a (virtual) parking lot attached, and an edge
has a specific value assigned to represent the distance between two blocks.

5.1.2 Parking Demand Modeling

In the simulation, we use the real-world traffic traces to generate the parking
demand. Here the parking demand is the number of drivers who need parking
spaces in the target area. However, in reality, it is difficult to collect the traffic traces
for parking in the target area. Although the sensor network is deployed to monitor
the incoming and outgoing traffic for parking in individual parking lot [16], [8],
the traffic data from individual parking lot cannot represent the total traffic traces
for parking in the whole area. Fortunately, we can employ the highway traffic
traces to estimate traffic for parking, which are available from the Performance
Measurement System (PeMS) at the University of California, Berkeley. Here we
make a general assumption that real total traffic for parking is proportional to the
highway traffic. Although not all of traffic pursuing parking spaces in target area
are from highways, and not all highway traffic need to park in the target area, the
highway case can simulate the state of total traffic for parking. We classify the
total highway traffic into incoming traffic and outgoing traffic, which represent the
traffic approaching to and leaving from the target area. The incoming traffic serves
 37

as the reference of parking demand. We use the driving distance within the target
area to measure the traffic for parking. For calculating the driving distance, the
vehicles begin to run the distance meters, once they enter the area, until reaching
the selected parking lot. Moreover, as we discuss in Section 3, we use congestion
factor l to represent the level of congestion condition. In the simulation, we use the
traffic load to simulate the congestion condition. For instance, if l = 1, the traffic
load is under normal condition. If l = 2, the traffic load is double the normal
condition.

5.1.3 Simulator Implementation

In order to implement different parking policies in the simulation environment,

we use object-oriented design to realize the interactions between objects (e.g.,
users and parking lots). Although there are thousands of autonomous drivers
who behave differently, they have the same common concerns about the parking
selection, including convenience degree and parking price. So the object-oriented
design provides us a cost-effective way to implement different behaviors and
common concerns of objects through structures within the program. Meanwhile,
by providing the interface of the objects, it is easy to extend the simulator to
other applications. Therefore, the simulator is developed not only for reservation
policy and dynamic pricing scheme, but to simulate general strategies of parking
selections.

5.2 Simulation Setup

In our simulation, we use the map of Los Angels Downtown as the target area, as
shown in Fig. 5.1, which is surrounded by the interstate highways, I-101, I-110, I-10
 38

Figure 5.1: Map of Los Angeles Downtown

and I5. I-n this area, there are multiple typical districts including central business
district, residential district, entertainment venues, which is one of busy areas in
Los Angeles. This area is very representative in big cities. Furthermore, Fig. 5.2
illustrates the incoming and outgoing traffic in two different days. As we see, the
peak time of incoming traffic is from 6am to 10am, and the rush hour of outgoing
traffic is during 5pm to 8pm. It matches people’s regular schedule, in the morning
most people drive to work and go back home after 5pm. Therefore, the traffic trace
is reasonable to generate the parking demand in the simulation. Furthermore, we
assume that the users’ budget distribution follows the Gaussian distribution at a
certain range from 1 to 10. If the budget is less than 1, we force it to be 1, and if it
is larger than 10, we set it to 10.
 39

Traffic on 05/01/09

14000 Incoming Traffic Flow

Outgoing Traffic Flow
12000

10000
Number of vehicles

8000

6000

4000

2000

0
0 5 10 15 20 25
Time (hour)

Traffic on 02/16/09
9000

8000

7000
Number of vehicles

6000

5000

4000

3000
Incoming Traffic
2000 Outgoing Traffic

1000
0 5 10 15 20 25
Time (hour)

Figure 5.2: Incoming and Outgoing Traffic Flows

 40

5.3 Experimental Results

The following experimental results illustrate the efficacy and feasibility of the
proposed Reservation-based Smart Parking (RSP) system in a cost-effective way.
We evaluate the effectiveness of reservation policy and dynamic pricing scheme in
terms of following perspectives:

5.3.1 Traffic Searching for Parking

Traffic Searching for Parking

4.5
Blind Search
Reservation Policy
PIS
4 BPIS
Average Driving Distance (mile)

3.5

2.5

1.5
0 5 10 15 20 25
Time (hour)

Figure 5.3: Traffic Searching for Parking Comparison under Different Parking
Guidance Strategies

As we discuss in Chapter 2, Fig. 5.3 shows that the driving distance under
blind search is the worst, especially during the peak hours; PIS and BPIS are better
than blind search when traffic flows increase; and the reservation policy is the best
 41

compared with others. Note that, in this simulation, there is no pricing scheme
implemented in reservation policy. An interesting observation of reservation
policy is that the average driving distance is decreasing at peak time, rather than
increasing. That is because, after users learn the states of parking lots, they tend to
reserve the nearest parking lot to their destination. During the peak hours, most
parking lots are almost fully occupied in central area. Consequently, users have
to select the parking lots in surrounding area, which are near to their start points.
Therefore, it results in the reduction of average driving distance during the peak
hours.
Related to the dynamic pricing scheme, fix-price policy is option for the service

Traffic Searching for Parking

1.92
Average Driving Distance under Dynamic Pricing
Average Driving Distance under Fixed Pricing
1.9

1.88
Average Driving Distance (Mile)

1.86

1.84

1.82

1.8

1.78

1.76

1.74
0 5 10 15 20 25
Time (Hour)

Figure 5.4: Traffic Searching for Parking under Different Pricing Schemes

provide. In order to compare the traffic searching for parking under these two
 42

different pricing options, we implement this simulation. In this simulation, both

fixed price and basic price of dynamic price are set as default value ($1 for per
time slot). For the dynamic pricing scheme, β u , β s , β c are set to 0.5, 0.3, 0.2, which
represent the different weights for different price components. For user’s utility
function, υ and κ are set to 1 and 0.3. As shown in Fig. 5.4, by implementing
the dynamic pricing scheme, the average driving distance of all users for parking
searching can be reduced. In general, a user under dynamic parking price will
select a parking lot that can provide the highest benefit. In contrast, under fixed
pricing, each user cannot learn the related parking information from the price and
is only concerned about the walking distance. It allows selfish users to pursue
the nearest parking lots to their destination without any limitation. As a result,
the parking lots busy districts are fully occupied, which force part of users to
drive more and waste their waste their gasoline. Contrarily, the occupancy rate in
surrounding districts is much lower. It also causes the problem of load balancing.
Fig. 5.4 shows that, during the initial periods, the performance of driving distance
is similar under two different schemes. Then, with the traffic load increasing, the
parking pricing is growing and the congestion price is charged to prevent further
congestion. In response, some users switch their selections from the high-priced
parking lots to others at cheap price. As the result, the parking demand is better
balanced across all parking lots, resulting in the reduction of driving distance,
compared with fixed-price scheme.

5.3.2 Revenue

We first compare the performance of revenue growth under proposed dynamic

pricing and fixed pricing, at different price units. Fig. 5.5 shows the total revenues
 43

5
x 10
6
Total Revenue under Dynamic Pricing
Total Revenue under Fixed Pricing
5

4
Revenue (Dollar)

0
1 2 3 4 5 6 7 8 9 10
Price Unit (Dollar)

Figure 5.5: Revenue Comparison between Dynamic and Fixed Pricing

of all parking lots in the target area according to different pricing schemes imple-
mented in reservation-based policy. Note that the price unit means basic price for
dynamic pricing scheme, and unique fixed price for fixed pricing scheme. Depend-
ing on different price units, both kinds of pricing perform differently on revenue
growth. Compared to fixed pricing, the proposed dynamic pricing can improve
the revenue when the basic price is small. With the increasing of basic price, the
revenue under dynamic pricing becomes similar with fixed price. Although the
largest revenue is lower than fixed price (94.22%), dynamic pricing can let more
valuable users with lower budgets find their desirable parking spaces. Since the
users’ budget constraints are under Gaussian distribution, higher basic price will
cause more failure on parking searching and traffic congestion. Therefore, in real
world, low price is more feasible considering the social welfare, which allows more
users with lower budget get their parking spaces. In this case, the dynamic pricing
 44

can achieve better revenue, because it can adjust parking price dynamically based
on parking condition, as well as let most users be satisfied.
Additionally, since setting the fixed price to obtain the most revenue is based the

5
x 10

Total Revenue under Dynamic Pricing

Total Revenue under Random Selection of Fixed Pricing
5

4
Revenue (Dollar)

0
3 [1ï5] 5 [1ï10] 4 [3ï6]
Price Unit (Dollar)

Figure 5.6: Revenue Comparison between Dynamic and Random Selection of Fixed
Pricing

distribution of users’ budgets, it is difficult to realize such distribution in the real

world. So random selection of fixed pricing is considered as a potential option
without the information of users’ budget. As shown in Fig. 5.6, the revenue under
random pricing depends on the random range. When the range selected is small
enough, the revenue is getting close to the fixed price. With the increasing of
random range, the performance of revenue growth becomes much worse than
proposed dynamic pricing and fixed pricing. Therefore, this experimentation
results show that the proposed dynamic pricing can adjust price according to real-
time data to obtain more revenue and satisfy more users than fixed and random
 45

selection pricing, when the basic price is small. However, when the selection range
is small enough, the provider obtains similar revenue under random selection and
fixed price, but sacrifice the users’ satisfaction.
The results in this simulation indicate that the proposed dynamic pricing scheme
uses the real-time parking condition as users’ feedback, takes advantage of users’
differentiated needs for the improvement in revenue growth, as well as concerns
users’ satisfaction, relative to the fixed-price and random-price policies.

5.3.3 Service Differentiation

0.9

0.8
Users with budget $2
0.7 Users with budget $5
Users with budget $10
0.6
Failure Rate

0.5

0.4

0.3

0.2

0.1

0
1 1.5 2 2.5 3 3.5 4 4.5 5
Congestion Factor

Figure 5.7: Failure Rates of Users with Different Budgets

The proposed dynamic pricing scheme can guarantee the service differentiation,
even if the traffic congestion happens. In this simulation, we only implement the
 46

dynamic-price scheme, keep the basic price at the default value (which is $1 for
per time slot), and vary the traffic load to check the failure rates of the users
with budget $2, budget $5 and budget $10. The failure rate here is defined as the
number of drivers who cannot find parking spaces versus the number of total
drivers. As shown in Fig. 5.7, all failure rates of these users is seen to increase
with the congestion factor, which is determined by traffic load in this simulation.
Initially, under the default basic price, all users can find their desired parking
spaces. With the load increasing, the failure rate of users with budget $2 soars
firstly, and then more users with budget $5 start to be unable to find parking spaces.
At the heavy load, the dramatic parking demand causes the higher failure rate of
users with budget $10. However, when the congestion happens, the dynamic-price
scheme can guarantee the quality of parking service for the users with higher
budgets, compared with the users with low budgets. Furthermore, by adjusting
the price dynamically, it stimulates users to moving to uncongested parking lots
via lower price to help balance the traffic load.
 47

Chapter 6

Conclusion and Future Work

6.1 Conclusion

In this thesis, we have developed a new Reservation-based Smart Parking (RSP)

system to optimize parking management. We have proposed a dynamic pricing
scheme for satisfying the different needs of drivers and service providers, which is
based on real-time parking information. The pricing scheme is integrated with the
proposed RSP system in which parking price is dynamically adjusted in response
to the relationship of demand and supple and congestion level. Upon receiving
parking prices, drivers make their reservations to maximize their benefits according
to the utility function. Based on the obtained results from our simulation study, we
conclude that the proposed RSP system increases the revenue for service providers,
provides service differentiation for users with different needs, alleviates traffic
congestion caused parking searching and reduces the amount of traffic searching
for parking.
 48

6.2 Future Work

In this section, we discuss future work of the model implementation.

6.2.1 Synchronization

In the proposed RSP system, sensing system is responsible for collecting parking
information periodically. Synchronization of the parking states in different parking
lots is required. Along this direction, the authors in literature [14] proposed a
fully localized diffusion-based method to achieve full scalability, and Jana et al [26]
introduced lightweight tree-based synchronization (LTS) algorithm to operate the
node failures. In the proposed system, the management system collects the parking
information from individual parking lots, and generates the parking guidance
signal (i.e., parking price). Hence, with the centralized management system, it is
trivial to achieve the clock-synchronization of the sensed data.

6.2.2 System Deployment

In the real world, it is impossible to make all of drivers adopt the proposed RSP
system at once. Increasing the deployability is an important concern. Fortunately,
by implementing the reservation policy and dynamic pricing scheme, the proposed
RSP system is in favor of the drivers using such system, when compared to users
without using the system. Hence, it provides enough incentive for drivers to adopt
the system.

6.2.3 Security and Privacy

One of the concerns in our proposed RSPS is the security of communication

and users privacy. Based on a secure symmetric encryption algorithm [17], all
 49

communication between service providers and users are encrypted. Additionally,

to protect users’ privacy, we may use virtual identifier for each user when the user
is authenticated in the system. As a result, the real identifiers of users are hidden
from the adversary.
 50

Bibliography

[1] Open spot. http://openspot.googlelabs.com/.

[2] Solutions for improving city operations. http://www.streetlinenetworks.

com/.

[3] The global warming survival guide. Time magazine, 2007.

[4] Transportation Energy Databook. U.S. Department of Energy, 2010.

[5] Ella M. Atkins. Cyber-physical aerospace: Challenges and future directions

in transportation and exploration systems. In Technical Report CMU/SEI-2011-
TR-002, 2011.

[6] R. Baheti and H. Gill. Cyber-physical systems. In The Impact of Control

Technology, 2011.

[7] Murat Caliskan, Daniel Graupner, and Martin Mauve. Decentralized discovery
of free parking places. In Proc. of the Third ACM International Workshop on
Vehicular Ad Hoc Networks (VANET’06), 2006.

[8] Jatuporn Chinrungrueng, Udomporn Sunantachaikul, and Satien Triamlum-

lerd. Smart parking: an application of optical wireless sensor network. In
Proceedings of Application and the Internet Workshops, 2007.
 51

[9] Edmund M. Clark, Bruce Krogh, Andre Platzer, and Raj Rajkumar. Analysis
and veri cation challenges for cyberphysical transportation systems. In NI-
TRD National Workshop for Research on Transportation Cyber-Physical Systems:
Automotive, Aviation, and Rail, 2008.

[10] Michal Feldman, Kevin Lai, and Li Zhang. A price-anticipating resource

allocation mechanism for distributed shared clusters. In Proceedings of the 6th
ACM conference on Electronic commerce, 2005.

[11] F.P.Kelly, A.K.Maulloo, and D.K.H.Tan. Rate control for communication

networks: shadow prices, proportional fairness and stability. Operational
Research Society, 49:237–252, 1998.

[12] Alain L. Kornhauser. About prospect eleven and the DARPA project. J. Robot.
Syst., 23(9):745–753, 2006.

[13] I. Lee and O. Sokolsky. Medical cyber physical systems. In Design Automation
Conference (DAC 2010), 2010.

[14] Qun Li and Daniela Rus. Global clock synchronization in sensor networks. In
Proceedings of IEEE INFOCOM’04, 2004.

[15] Todd Litman. Smart transportation emission reduction strategies. 2011.

http://www.vtpi.org/ster.pdf.

[16] Rongxing Lu, Xiaodong Lin, Haojin Zhu, and Xuemin (Sherman) Shen.
SPARK: A new VANET-based smart parking scheme for large parking lots.
In Proceedings of IEEE INFOCOM’07, 2007.

[17] Wenbo Mao. Modern Cryptography: Theory and Practice. Prentice Hall, 2003.
 52

[18] Suhas Mathur, Tong Jin, Nikhil Kasturirangan, Janani Chandrashekharan,

Wenzhi Xue, Marco Gruteser, and Wade Trappe. ParkNet: Drive-by sensing
of road-side parking statistics. In MobiSys’10, 2010.

[19] B. McMillin, C. Gill, M. L. Crow, F. Liu, D. Niehaus, A. Potthast, and D. Tauritz.

Cyber-physical systems distributed control: the advanced electric power grid.
Electrical Energy Storage Applications And Technologies, 2007.

[20] O’Sullivan and Steven M. Sheffrin. Economics: Principles in action. Pearson

Prentice Hall, 2006.

[21] Ragunathan (Raj) Rajkumar, Insup Lee, Lui Sha, and John Stankovic. Cyber-
physical systems: The next computing revolution. In Design Automation
Conference, 2010.

[22] Jake Reisdorff, Aldo Arizmendi, and Jason Armstrong. Parking lot detection
proposal.

[23] Krishna Sampigethaya, Radha Poovendran, and Linda Bushnell. Secure

operation, control and maintenance of future e-enabled airplanes. IEEE special
issue on Aviation Information Systems, 96(12):1992–2007, December 2008.

[24] Susan A. Shaheen and Caroline J. Rodier. Smart parking management field
test: A bay area rapid transit (BART) district parking demonstration. UCD—
ITS—RR—05—02, 2005.

[25] Donald C. Shoup. Cruising for parking. Transport Policy, 13:479–486, 2006.

[26] Jana van Greunen and Jan Rabaey. Lightweight time synchronization for
sensor networks. In Proceeding ACM WSNA’03, 2003.

[27] Hal R. Varian. Microeconomic Analysis. New York: Norton, 2003.

 53

[28] Xin Wang and Henning Schulzrinne. Pricing network resources for adaptive
application. IEEE Transactions on Networking, 2005.

[29] Paul Steely White. No vacancy: Park slope’s parking problem. Transportation
alternatives, 2007. http://www.transalt.org/newsroom/releases/126.

[30] Wayne Wolf. The good news and the bad news. IEEE Computer,
10.1109/MC.2007.404, 2007.

[31] Daniel Work, Alexandre Bayen, and Quinn Jacobso. Automotive cyber phys-
ical systems in the context of human mobility. In National Workshop on
High-Confidence Automotive Cyber-Physical Systems, 2008.

[32] Daniel B. Work and Alexandre M. Bayen. Impacts of the mobile internet
on transportation cyberphysical systems: Traffic monitoring using smart-
phones. In National Workshop for Research on High-Confidence Transportation
Cyber-Physical Systems: Automotive, Aviation and Rail, 2008.