International Journal of Computer Networking,

Wireless and Mobile Communications (IJCNWMC)

ISSN (P): 2250-1568; ISSN (E): 2278-9448
Vol. 7, Issue 4, Dec 2017, 1-10
© TJPRC Pvt. Ltd.

THE OPTIMAL CARPOOL PLANNING, BASED ON IOS PLATFORM

HER-SHING WANG, CHUN-TZE HSIAO & WEI-CHEN DONG

Department of Industrial Engineer and Management, National Taipei University of
Technology, 1, Sec.3, Zhongxiao E. Rd., Taipei, Taiwan
ABSTRACT

The rapid development of technology has aggravated global warming and air pollution, particularly in
metropolises, with high population density. Urban overpopulation has also caused serious traffic problems. The area
around the Neihu Technology Park in Taipei City, for example, has long suffered from traffic congestion, during peak
hours. To address such urban traffic issues, innovative transportation services are as important as modification, in design
of major roads. Taipei City Government, therefore, intends to implement traffic control in favor of high-occupancy
vehicles and promote taxi-based carpools so as to alleviate traffic jams. In light of the above, this study focuses on taxi
sharing services and on how to make the services more convenient and more flexible by using a mobile application
program (or APP for short) supported by the IOS platform. Based on each passengers boarding and alighting locations,
which are input into the APP, the APP performs passenger matching and route planning via cloud computing and the
NSGA-II algorithm, in order for a taxi driver to view passenger information and the planned shortest route, and for each

Original Article
passenger to know the driver’s information, the route, the vehicle departure time, the estimated arrival time, the fee, and
so on. Working on mutual trust, this system features transparency, real-time operation, and traceable riding information
to help increase the use of carpools and add more value to vehicle sharing services.

KEYWORDS: Taxi Sharing, IOS, Cloud Computing, NSGA-II Algorithm & Route Planning

Received: Sep 18, 2017; Accepted: Oct 08, 2017; Published: Oct 28, 2017; Paper Id.: IJCNWMCDEC20171

INTRODUCTION
Background and Motivation of the Study

Global warming and air pollution are exacerbated by the rapid development of technology, especially in
highly populated urban areas. According to data published by the Directorate- General of Budget, Accounting and
Statistics (DGBAS) of the Republic of China, Taiwan ranks second in the world in population density and gasoline
consumption per capita while having the highest vehicle density. One major source of air pollution in Taiwan,
therefore, is the pollutants emitted from cars and motorcycles, and the reduction of vehicle exhaust, a critical step to
prevent and control environmental pollution. Currently, vehicle occupancy is low in urban areas, resulting in not
only traffic congestion but also severe air pollution. If carpools can be effectively promoted, with properly planned
routes, and with passengers’ needs taken into consideration, more and more people will be willing to participate,
thus reducing the number of cars on the road and mitigating air pollution.

Early vehicle sharing services in Taiwan were carried out as follows. Vehicle sharing needs were typically
posted on bulletin boards of a local area, community, or company, and those who were willing to share a vehicle
with others registered themselves by phone and made arrangements for the pickup time and places. Nowadays, the
rapid advancement of smart phones and tablet computers, together with the rise of intelligent transportation

www.tjprc.org [email protected]
2 Her-Shing Wang, Chun-Tze Hsiao & Wei-Chen Dong

systems, has made possible a vehicle sharing system that incorporates such technologies as vehicle positioning, wireless
communication, and dynamic real-time matching, and that allows only a member of the system to log in to the system
through a mobile phone as a way to solve security problems involving personal riding habits and privacy. Moreover, in
place of the traditional posts, automatic matching mechanisms have emerged to match carpool passengers by means of
web-based algorithms, which greatly enhance the flexibility and convenience of vehicle sharing services.

To encourage more people to use vehicle sharing services, it is important to satisfy passengers’ needs and provide
door-to-door services through a convenient and easy-to-use system. The system should be able to eliminate the
conventional inconveniences of matching passengers by the passengers themselves and perform automatic matching to not
only reduce the time required to look for a vehicle and other passengers, but also screen vehicles and passengers according
to users’ needs. Furthermore, the system should allow a user to check out routes, vehicle information, and so on, with a
view to reducing the user’s sense of insecurity during a shared ride

Objectives of the Study

This study aims to establish a mathematical model for the problem stated above and to develop an algorithm for
solving the problem, taking into account the needs of urban passengers in particular. Given the emphasis an urban
passenger tends to place on time, two major factors that influence an urban inhabitant’s willingness to use a vehicle sharing
system are the thoughtfulness of the system and the ride time. While sharing vehicles allows passengers’ travel needs to be
advantageously satisfied with the smallest number of vehicles and traffic jams to be relieved at the same time, travel time
must also be shortened to add to the appeal and value of vehicle sharing. In addition, despite the advantages of vehicle
sharing, willingness to share a vehicle with others may be weakened by difficulties in matching passengers’ pickup time
and locations, in distributing fees among passengers, and in discounting the fees. It is hence imperative to know the factors
affecting the demand for vehicle sharing services and to provide an efficient and user-friendly system for assisting, and
ensuring long-term operation of, such services. This study begins by establishing a mathematical model based on urban
passengers’ vehicle sharing demands. Then, a multi-objective genetic algorithm is applied for solutions. An APP that
integrates both vehicle information and passenger information is thus developed. After a driver inputs a starting location, a
destination, and a departure time, and by connecting to a database and cloud computing, the APP matches passengers
wishing to share the vehicle, screening out passengers in the database whose demands cannot be satisfied by the current
conditions. Then, the APP plans an optimal route and sends the route to the driver’s and the matched passengers’ mobile
devices as an iOS map. The entire system is sound and stable and helps increase the public’s willingness to share vehicles
so that the number of vehicles on the road can be reduced to effectively prevent traffic jams.

The Objectives of this Study are

• To apply system development techniques to the creation of an iOS-based carpool matching service system, in
response to the government’s initiative to implement traffic control, in favor of high-occupancy vehicles and to
promote carpools in Neihu Technology Park, with a view to providing passengers, with a new mode of vehicle
sharing services; and

• To connect an APP to a database through the system platform; to match passengers by computing with a heuristic
algorithm to shorten passengers’ waiting time; and to perform proper taxi planning in order to pick up and
transport all the passengers to their destinations with the smallest number of vehicles and the shortest total travel

Impact Factor (JCC): 7.8476 NAAS Rating: 3.71

The Optimal Carpool Planning Based on IOS Platform 3

time, to prevent traffic congestion associated with an excessive number of vehicles, and to arrive at reasonable
fees for the benefit of both drivers and passengers.

To provide passengers with access to vehicle sharing information such as taxi drivers’ information, the
identification number of the vehicle to be shared, the passengers to share the vehicle with, the fee, and the planned route,
allowing passengers to have peace of mind while enjoying the convenient vehicle sharing services.

LITERATURE REVIEW
Discussion on Literature Involving Vehicle Routing Problems with a Time Window

After years of development, vehicle routing problems (VRP) have evolved into many variations, a notable
example of which is vehicle routing problems with a time window (VRPTW). The concept of time windows is important to
researches on VRP because in actual route planning, those who share a vehicle have different ride times respectively. While
most vehicle routing problems without time window constraints have been properly studied, VRP variations with such
constraints are still faced with challenges (Miranda and Conceição, 2016). The random travel time in VRPTW is more
difficult to determine than that in vehicle routing problems without a time window. Depending on the strictness of temporal
constraints, time windows can be divided into hard time windows and soft ones. A hard time window requires vehicle
arrival within a specified range of time and does not allow a vehicle to be late, for the services of the vehicle will be denied
if the vehicle arrives late. A soft time window, on the other hand, allows a late arrival to be dealt with via waiting or a fine.
Difficulties arise when the probability distribution of customer arrival time varies under the constraint of a hard time
window (Gendreau, Jabali, and Rei, 2014).

With the changes of industry types, VRPTW has been more widely applied and discussed than before, including
those involving security patrol services, mail delivery, school bus, and so on (Desaulniers et al., 2014). To solve such
problems effectively, Zhang et al. (2016) established three models, with estimated problem-solving time derived from
different perspectives and random demands. The first objective is to look for the delivery route of the lowest expected total
cost. The second objective is to maximize the probability of on-time delivery to all customers. The third objective is to
lower the expected total cost while ensuring that the time of delivery to each customer is within a specified time range.
The authors proposed and discussed two methods for solving problems with the three models. The first method is
preventive purchase, and the second is to detour to a warehouse in order to load a vehicle. According to the authors’ study,
the first method yields better solutions than the second does.

It can be known from the above that recent years have seen more and more emphasis placed on time window
constraints. With time being at a premium nowadays, it is highly desirable for delivery companies to reach each destination
rapidly and for customers to receive their purchased products as soon as possible. The ultimate goal is not only to shorten
travel distances and save energy, but also to contribute to sustainable development of the environment.

Multi-Objective Optimization

In real-world applications, optimization with a single objective is insufficient in dealing with decision problems
because the optimal solution tends to be compromised in quality, if not rendered infeasible, by any disturbance or change
of the input data. Solving a shortest-path problem, for example, requires a balance between time and cost, too, and the
objective functions involved are typically conflicting. Single-objective optimization is different from multi-objective
optimization mainly in that the former is unable to consider several conflicting objectives at the same time. This explains

www.tjprc.org [email protected]
4 Her-Shing Wang, Chun-Tze Hsiao & Wei-Chen Dong

why most problems in the real world need to be solved with multiple objectives in mind and, under the condition of
meeting multiple constraints, will have a Pareto-optimal solution as a non-dominated solution. A multi-objective
optimization problem involves a plurality of objective functions to be minimized or maximized. Mathematically, multi-
objective optimization can be expressed by the following equations and inequalities (Marko Kovačević et al., 2014), in
which: equation 2.1 denotes a multi-objective optimization problem, with k objective functions; gj(x) in inequality 2.2 is a
constraint function of inequality; hi (x) in equation 2.3 is a constraint function of equality; and xL and xU in inequality 2.4
are the lower limit and the upper limit of the variable, respectively.

Minimizef (x) = {f1 (x), f2 (x), …, fk(x)} (2.1)

s.t.gj(x) ≤ 0, j = 1, 2, …, J(2.2)

hi(x) = 0, i = 1, 2, …, I(2.3)

xL ≤ x ≤ xU (2.4)

As multi-objective algorithms can better simulate real-world problems, than their single-objective counterparts,
more and more scholars have begun devoting themselves to design and research in the former field. Schaffer is generally
recognized as the first scholar in the 1980s to propose the concept of the Multi-Objective Optimization Evolutionary
Algorithm (MOEA). Schaffer’s method is called the Vector-Evaluated Genetic Algorithm (VEGA), which adds a selection
mechanism to the conventional GA. VEGA, however, is flawed in many ways, the most important one of which is that the
solution obtained does not stay in an acceptable range and is not particularly prominent for any of the objective functions,
even if the solution is greater than the common average (D. Martín et al., 2014). Following VEGA, more multi-objective
algorithms were proposed. Deb et al. (2002) put forward the improved Non-Dominated Sorting Genetic Algorithm-II
(NSGA-II), which uses such steps as population sorting, crowding distance calculation, and the elitism strategy to
overcome slow convergence and the problem of being trapped in a local solution. Jenson (2003) reduces the complexity of
NSGA-II, by means of a different sorting process and a different data structure. Yijie and Gongzhang (2008), proposed
three modes of crossover operators, namely the maximum distance, the maximum-minimum distance, and the neighboring
maximum distance. Bo Feng et al. (2010) improved NSGA-II, by sorting and by using a particular chromosome selection
method. Reiter and Gutjahr (2012), incorporated NSGA-II with GA to solve dual-objective vehicle routing problems, with
capacity limitation, and the resulting Pareto solution set is somewhat enhanced in comparison with those obtainable by the
conventional approaches. Taking the foregoing into account, this study uses the number of vehicles and time, as factors in
shared-vehicle route planning and solves routing problems via NSGA-II, which according to the above is highly
applicable, and is efficient in finding solutions, to multi-objective problems.

MODEL ESTABLISHMENT
Definition and Hypotheses of the Problem to be Solved

The numbers of cars and motorcycles keep increasing in Taiwan, whose vehicle density has long been one of the
highest in the world. The urban areas in particular, such as the area around the Neihu Technology Park in Taipei City, are
notorious for traffic jams during peak hours. Apart from modifying the plans and designs of major roads, it is important to
find novel transportation service modes, as solutions to urban traffic problems. In this study, therefore, a vehicle sharing
service system, based on open vehicle routing problems is established by incorporating the use of handheld mobile devices
by taxis drivers and passengers, in order to provide multipoint-to-multipoint services from door to door. Passengers are

Impact Factor (JCC): 7.8476 NAAS Rating: 3.71

The Optimal Carpool Planning Based on IOS Platform 5

required to input their boarding locations and alighting locations in advance; the system will determine the corresponding
longitudes and latitudes through the Global Positioning System (GPS). Then, the system plans routes by distributing
vehicles and matching passengers. Both drivers and passengers can view the route planning results and other related
information via an APP, so that both parties benefit from the convenient services.

Limitations and Hypotheses of this Study are as Follows

• The coordinates of the starting location of each taxi are known.

• The quantity of passenger demands, (i.e., the number of passenger demand points) is known, and so are the
coordinates of all the boarding locations and alighting locations.

• Vehicles are divided into large vehicles and small ones, both having occupancy limitations.

• Passengers who are matched by the system to take the same vehicle are not allowed to deny the matching result
for any reason.

• All the vehicles run at the same speed and basically follow the upper speed limit of each road, as specified in the
map information.

• No vehicles need to return to their respective starting locations after completing the services, so only one-way
travels are planned.

• It is assumed that there is no upper limit on the number of vehicles available for services.

Mathematical Model

The route planning problems to be solved by the taxi-sharing passenger matching service system in this study are
similar to open routing problems. Objective functions for achieving the smallest number of vehicles and the shortest total
time are used to establish the mathematical model of this study. All the symbols in the model and their definitions are listed
in Table 1.

Table 1: Symbol Definition

Symbol Definition
Parameter G Network Diagram, G = (V, A)
N A set of nodes, D = {d0, d1, …, dn}, in which node d0 is the starting location
A A set of nodal lines, A = {(vi, vj): v0, v1∈N, i≠j, j≠ 0}
V A set of taxis, V = {1, 2, …, v}
K K∈ {1, …, m}, denoting a set of m different vehicles types
k k∈K, denoting vehicle type k
Qk Denoting the upper occupancy limit of a taxi of vehicle type k
Denoting the number of passengers to share a vehicle with that is acceptable to the ith
CQi
passenger
qi Indicating that passenger demand point i has a non-negative demand
n Denoting the number of passenger demand points
Denoting the time it takes for the vth taxi to travel from node i to node j
Denoting the time at which the vth taxi begins services at node i
Li Denoting the lower limit of the time window at node i
Ui Denoting the upper limit of the time window at node i
Denoting the time at which the vth taxi arrives at node i

www.tjprc.org [email protected]
6 Her-Shing Wang, Chun-Tze Hsiao & Wei-Chen Dong

Table 1: Contd.,
Denoting the time at which the vth taxi departs from node i
1, the taxi of vehicle type travels from node $ to node %(
Decision variables
0, otherwise
1, node $ is assigned to the vehicle(
)
0, otherwise

Target Functions

(3.1)

(3.2)

Constraint Functions

(3.3)

(3.4)

∀ = 1, 2, … ,
∀5 = 1, 2, … , 6

(3.5)

(3.6)

Impact Factor (JCC): 7.8476 NAAS Rating: 3.71

The Optimal Carpool Planning Based on IOS Platform 7

(3.7)

(3.8)

(3.9)

(3.10)

(3.11)

(3.12)

(3.13)

(3.14)

www.tjprc.org [email protected]
8 Her-Shing Wang, Chun-Tze Hsiao & Wei-Chen Dong

Function Description

(3.1) Target function 1 minimizes the total number of vehicles.

(3.2) Target function 2 minimizes the total amount of time.
Indicating that each demand point can be serviced by only one taxi, which after
(3.3), (3.4) completing its services can leave directly and need not return to its starting
location.
(3.5) Indicating that the flow rate at each demand point stays constant.
Indicating that the total number of demand points in each route cannot exceed
(3.6)
the upper occupancy limit of a taxi of vehicle type k.
Indicating that the total number of demand points in each route cannot exceed
(3.7) the number of passengers to share a vehicle with that is acceptable to the ith
passenger.
Indicating that each vehicle leaves immediately and will not return to its starting
(3.8), (3.9)
location, after completing its services.
To ensure that the time at which the vth taxi departs from node i exceeds the
(3.10) upper limit of the time window at node i, with Max being a positive integer
maximum.
To ensure that the time at which the vth taxi arrives at node i does not exceed the
(3.11) lower limit of the time window at node i, with Max being a positive integer
maximum.
To ensure that the time at which the vth taxi departs from node i exceeds the time
(3.12) at which the kth taxi arrives at node i and the time at which the kth taxi begins
services.
To ensure that the time at which the vth taxi arrives at node i exceeds the time at
(3.13) which the kth taxi departs from node i and the time it takes for the kth taxi to
travel from node i to node j
(3.14) To set the range of the service time of the vth taxi at node i

Practical System Operation and Case Analysis

To ensure that, the route planning algorithm of the APP in this study can find the optimal route, which features the
smallest total number of vehicles and the shortest driving time, system simulation experiments were carried out at
“different scales” and with “different percentages of passengers willing to share vehicles with others”. The goal was to
minimize the number of vehicles and achieve the shortest total travel time. At the scale of 50 passengers, 30 random
experiments were conducted, each with the percentages of passengers willing to share vehicles with others being 0%, 25%,
50%, 75%, and 100%. The NSGA-II algorithm was used to obtain the optimal solutions, and single-factor analysis of
variance (ANOVA), to determine whether an increase in the level of vehicle-sharing willingness has a significant effect on
attaining the objective of this study with the vehicle-sharing passenger matching and route planning system.

As shown in Table 2, the case scale of this study was set at 50 passengers, and the percentages of passengers
willing to share vehicles with others were set at 0%, 25%, 50%, 75%, and 100%. The matching of vehicle-sharing
passengers and the route planning were performed without violating the condition that passengers’ time windows as well as
the limitation on the number of passengers sharing a vehicle must be respected. The longitudes and latitudes of passenger
demand points were generated randomly.

Impact Factor (JCC): 7.8476 NAAS Rating: 3.71

The Optimal Carpool Planning Based on IOS Platform 9

Table 2: Case Scale

Scale 50 Passengers
0%
25%
Percentage of passengers willing to share vehicles with others 50%
75%
100%
Number of experiments 50

The results of 50 random experiments were plotted as Figure 1 and Figure 2.

Figure 1: Broken line Graph of Total Number of Vehicles for 50 Passengers, with
Different Percentages of Passengers Willing to Share Vehicles with Others

Figure 2: Broken Line Graph of Total Travel Time of 50 Passengers, with Different
Percentages of Passengers Willing to Share Vehicles with Others

The experiment results show that, given different levels of vehicle-sharing willingness, higher levels of
willingness tend to reduce the number of vehicles and the travel time more. Besides, according to the results of ANOVA,
with α = 0.05, the p-value of significance = 0.00 (< α = 0.05), meaning there are significant differences in the total numbers
of vehicles and in the total travel time at the same passenger scale and with different levels of willingness to share vehicles.

CONCLUSION AND SUGGESTIONS

To effectively promote vehicle sharing services, these services must be able to save the time cost of vehicles while

www.tjprc.org [email protected]
10 Her-Shing Wang, Chun-Tze Hsiao & Wei-Chen Dong

reducing traffic congestion. This study proposes an iOS-based vehicle-sharing passenger matching service system APP that
integrates both vehicle information and passenger information. The APP can connect to a rear-end database, through a
mobile network to enable rapid data transmission, compilation, and computation. The APP takes advantage of the GPS
function of mobile devices, competes with the NSGA-II algorithm, and presents the planned routes via maps built in iOS,
so that system users can transmit and receive information in real time. This study has the following achievements:

• The establishment of an APP system: The APP in this study has a front-end user interface created with Xcode,
stores data in a MySQL database, and ensures connection between the user interface and the database. Operation
of the APP system begins by computing passengers’ demands, in order to match passengers. Once a route is
planned, the cloud computing results and other related data are sent back to passengers’ mobile devices through
mobile networks in order to show the route in an iOS map, along with other related information.

• The matching of vehicle-sharing passengers and the planning of routes: This study aims to achieve the smallest
number of vehicles and the shortest travel time, by computing with a heuristic algorithm without violating
passenger-imposed constraints. To start with, passengers are matched to shorten their waiting time. Then, a route
is planned for the passengers in a multipoint-to-multipoint manner. Experiment results show a significant
reduction in the number of vehicles needed for vehicle sharing services and in travel time.

REFERENCES

1. Bo Feng, Zhong-Zhong Jiang, Zhi-Ping Fan a, Na Fu, “A method for member selection of cross-functional teams using the
individual and collaborative performances”, European Journal of Operational Research, vol. 203, 2010, pp.652-661.

2. Douglas Moura Miranda, Samual Vieira Conceição, “The vehicle routing problem with hard time windows and stochastic
travel and service time”, Expert Syst. Appl., vol. 64, 2016, pp. 104-116.

3. Marko Kovačević, MilošMadić, Miroslav Radovanović, DejanRančić, “Software prototype for solving multi-objective
machining optimization problems: Application in non-conventional machining processes’’, Expert Systems with Applications,
vol.41, 2014, pp. 5657-5668.

4. Peter Reiter, Walter J. Gutjahr, “Exact hybrid algorithms for solving a bi-objective vehicle routing problem’’, Central
European Journal of Operations Research, vol. 20, iss.1, 2012, pp.19-43.

5. Sun Yijie, Shen Gongzhang, “Improved NSGA-II Multi-objective Genetic Algorithm Based on Hybridization-encouraged
Mechanism’’, Chinese Journal of Aeronautics, vol. 21, 2008, pp. 540-549.

Impact Factor (JCC): 7.8476 NAAS Rating: 3.71