Anda di halaman 1dari 8

2016 IEEE 30th International Conference on Advanced Information Networking and Applications

An Architecture for Monitoring and Improving


Public Transportation Systems
Pedro H. S. Duarte∗ , Luis F. Faina∗ , Lásaro J. Camargos∗ ,
Luciano B. de Paula† and Rafael Pasquini∗
∗ Faculty of Computing (FACOM/UFU), Uberlândia – MG – Brazil

Email: pedrohsd@mestrado.ufu.br, {luisfaina, lasaro, rafael.pasquini}@ufu.br


† Federal Institute of Education, Science and Technology of São Paulo (IFSP), Bragança Paulista – SP – Brazil

Email: lbernardes@ifsp.edu.br

Abstract—Brazilian public transportation systems are facing adapting the required infrastructure in the cities, exacerbated
a significant demand reduction, mainly due to the poor quality common problems in large urban centers, such as traffic jams,
of the offered services, lack of information regarding lines and pollution, number of accidents and urban inequality [3].
timetables, high cost and lack of investment from the government.
Even though it is not trivial to improve financial aspects related
Specifically to the public transportation system, data col-
to the public transportation system, this work claims that the lection is extremely important in order to design and improve
overall system quality can be improved through ubiquitous the service. In Brazil, the existence of inspectors in charge of
data collection according to a proposed ontology, which is the observing the public transportation system is common, gener-
basis for knowledge extraction to support the required quality ating all the (non-ubiquitous) data that companies responsible
of experience improvements. The proposed architecture relies
on standard technologies available nowadays, providing a low
for the transportation systems have in order to design the
cost solution for the required data collection and analysis. entire transport network. So, in most of the Brazilian cities,
This paper presents the proposed inter-networking architecture an economically interesting (standard-technologies-based) so-
and ontology, then evaluates the system performance using a lution for automated and ubiquitous data gathering from
prototype developed with standard solutions. the transportation system is extremely desirable, which can
Index Terms—Internet of Things (IoT), Machine-to-Machine
significantly improve the quality of the information collected,
(M2M), Ontologies, Public Transportation Systems, Smart
World, Ubiquitous Computing. and also improve the overall transport network design.
This work leverages the M2M (Machine-to-Machine) and
IoT (Internet of Things) scenarios to propose an inter-
I. I NTRODUCTION
networking architecture aimed at improving the data gathering
In the last few years the demand over the public transporta- process in Public Transportation Systems. At this moment, our
tion systems in Brazil has been significantly reduced, while focus is the Brazilian Public Transportation Systems based on
the average amount spent by Brazilian families with private, buses, and the main objective is to automate the collection
mostly individual, transportation has increased [1], reaching of the maximum as possible statistics from the transportation
five times the average amount spent with public transportation system and to store such information in a cloud-based Public
[2]. The competitiveness loss observed in the Brazilian public Transport Data Service solution, according to a proposed
transportation system was caused by many factors, including ontology, in order to allow running analytic processes on it
the poor quality of the service offered, its high cost, lack for achieving knowledge-based improvements required in the
of information regarding lines and timetables and also lack overall transportation system.
governmental investment [3]. This paper details the proposed inter-networking architec-
At the same time, a globally observed phenomenon indicates ture and the proposed ontology. Ontologies are explored in
that from 2007 more than 50% of the world population is liv- various areas and studies, due to its ability to promote the
ing in cities, moving from rural areas to metropolitan regions sharing of knowledge bases and inter-operability between
[4]. Consequently, the infrastructure of cities needs to adapt systems. Applying an ontology, we firstly expect to extract
in order to support this population growth. However, such the required knowledge in order to improve the public trans-
adaptations are not straightforward in most of the Brazilian portation system quality, providing better quality of experience
cities, since the migration usually occurs before the required to the users.
infrastructure expansions. Afterwards, as another outcome of this proposal (inter-
Moreover, the raise of Brazilian population income observed networking architecture and ontology), we consider that many
in recent years, combined with government incentives for some auxiliary tools can be created. For instance, drivers could be
industrial sectors, including the entire automotive industry, sig- aided by buses’ embedded systems that provide the timetable
nificantly reducing tax on such industrial sectors, contributed to follow. Passengers at bus stops could receive buses ETA
to part of the population opting for the private transportation. (Estimated Time of Arrival), total number of seats and current
Such increase in the number of vehicles and the difficulties for load. The entire transport network could be automatically

1550-445X/16 $31.00 © 2016 IEEE 871


DOI 10.1109/AINA.2016.39
adjusted based on knowledge extracted from the data collected. • Security and public safety: such as video surveillance,
In this way, we hope that public transport become more data analysis, and work-flows for emergency services;
attractive, by providing easily accessible information, and • Smart buildings: which include smart meters, light, heat-
more efficient, by avoiding delays of buses and identifying ing, internal security systems, and water and waste man-
possible improvements in the system routes. agement.
Besides the proposed inter-networking architecture and the The context of this paper lies in the smart transportation
proposed ontology, another contribution of this paper is the scenario. There are several projects on this context as follows.
developed prototype using standard solutions available in the Libelium1 , an enterprise which produces an open source
market. The prototype shows the feasibility of collecting sensor platform, has tested a system of sensor platforms that
data using standard technologies, and allowed us to presents will measure the availability of parked vehicles. As part of
results on the load generated by the messages collected in the European Framework 7 Program, their monitoring system
the proposed scenario and the required to move the collected includes magnetic sensors, signal conditioning electronics, 7-
information towards the cloud-based Public Transport Data 10 year life battery and a radio. The measured information is
Service storage. transmitted to an access point then to the cloud, where the
The remainder of this paper is organized as follows. Section processed data is used by the parking department headquar-
II presents some background literature relevant to this work, ters to send to displays on the street as well as to mobile
introducing current projects in the M2M and IoT scenarios. phones/tablets/computers to direct vehicles to the appropriate
Section III details the proposed inter-networking architecture, available parking spots. Besides that, there is a node system
explaining its main components, and the proposed ontology. on lamp posts that senses the air quality as well as light level
Section IV presents the evaluations performed with our proto- to control the ambient light.
type, the information regarding the overall system performance Another similar parking system, called SF Park2 , was in-
and overhead for data gathering, and based on random data stalled in San Francisco. The system determines the status of
generation over the prototype, this section also briefly intro- parking spaces and informs individuals of their existence via
duces some analyses doable using the collected data. Section mobile phones/tablets. Additionally, it provides information to
V concludes this paper and describes the future work. traffic enforcement when a parking meter is expired as well
as enables parking availability based prices.
II. BACKGROUND A Romanian tech firm called Onyx Beacon3 is teaming up
with the local authorities in Bucharest to install beacons on
The world is going from a Machine-to-Machine model
buses to inform the people about the buses. The project aims to
(M2M) to a new paradigm called Internet of Things (IoT)
the visually impaired people that, using a smartphone, receives
[5]. Both paradigms have risen from the new technological
audio information when a bus is near the stop. This smart
solutions of the last decades, including the decreasing costs
system frees visually impaired users from having to rely on
of devices and the broad adoption of the Internet. The M2M
another person to tell them when their bus has turned up. A
model is built around solutions for specific problems. Basi-
smartphone app is used, which receives signal informing the
cally, an M2M solution is based on a device, which senses
app which bus is arriving so the passengers can check if that
some characteristic of an asset, and an application, which
is the one they are waiting.
consumes this data. The application is self contained and any
In some European cities, instead of displays, “smart tags”
intent to integrate different applications and share data is not
are being embedded in bus station walls, tourist landmarks,
simple.
and airport arrivals lounges, allowing citizens to access in-
An evolution of such environment is the IoT. The idea
formation using their smartphones. In Strasbourg, users can
is to integrate all the devices, systems, sensors to a broader
see if a bus is going to be late, and whether there is a bike-
scenario. In an IoT, not just data generated and consumed by
share service nearby. The system shows how many bikes are
people are available in Internet, but also data from real world
available and the distance from there4 .
objects [5]. A scenario which IoT may be explored is the
In the City of Cologne, Germany, an IBM pilot project was
one called Smart Cities, or more widely called Smart World.
used to predict and manage traffic flow and road congestion.
Nowadays the major part of the people lives in urban areas,
The pilot demonstrated how the city of Cologne can anticipate,
creating a demand for smarter services. In [5] it is cited many
better manage, and in many cases, avoid traffic jams and
types of so called “smart services”, including:
trouble spots across the city using analytic technologies. The
• Smart transportation: e. g., solutions for parking, traffic city’s traffic engineers and IBM were able to predict traffic
monitoring, public transport and municipal fleet manage- volume and flow with over 90 percent accuracy up to 30
ment; minutes in advance. As a result, travelers would be able to
• Smart health-care: which includes management of elec-
1 Libelim - www.libelium.com (accessed on Sept. 12, 2015)
tronic records, hospital asset management, and remote 2 SFPark - www.sfpark.org (accessed on Sept. 12, 2015)
monitoring systems; 3 Onyx Beacon - http://www.onyxbeacon.com/ (accessed on 13 Sept., 2015)
• Smart education: including e-Learning and connected 4 Fast Coexist - http://www.fastcoexist.com/1681809/smart-displays-beam-
campuses; city-information-directly-to-your-phone (accessed on Sept. 13, 2015)

872
better plan ahead and determine whether they should leave As soon as bus B1 approaches the stop point SP1 and
at a different time, plan an alternate route or use a different becomes covered by the wireless signal of the access point
mode of transportation. The traffic command center collects AP1 available at SP1 , the time-window ΔT{B1 ,SP1 } available
real-time data from more than 150 monitoring stations and 20 for the communication process is open. The time-window is
traffic cameras on the roads, highways and at intersections that available until the instant whenever B1 departs from SP1 and
are known to be traffic hot spots5 . the connection to AP1 is lost due to being uncovered by the
The work presented in this paper lies in the context of wireless signal. The ideal departing time is the instant defined
those others presented in this section (and many others that in the timetable, but fluctuations can occur due to, for example,
may be found in the literature), and as such, it is designed elevated number of embark and disembark of passengers.
to be an IoT solution, in which the proposed ontology is In general, every time a given bus Bx arrives at a given
expected to provide the basis for knowledge exchange among stop point SPi , the two main actions that can occur is the
different solutions, a step further from a simpler M2M scenario embark and the disembark of passengers. Obviously, all the
with a single aim of improving only the bus-based public combinations of such two actions are possible, including no
transportation system. In the next section, the design of the embark and no disembark of passengers. As higher the number
solution proposed in this work is defined. of passengers for each action at the stop point, the higher the
amount of time expected for bus Bx being stopped at stop
III. P ROPOSAL point SPi . So, the time-window ΔT{Bx ,SPi } is in function of
In this section firstly the proposed inter-networking architec- the wireless coverage area of the access point APi available at
ture responsible for the ubiquitous data collection is detailed. SPi , the number of passengers embarking at SPi , the number
This architecture collects data from the public transportation of passengers disembarking at SPi , and some extra time due
system and pushes it towards the data repository at a Public to unexpected occurrences i . Figure 2 depicts the sequence
Transport Data Service. Afterwards, in this section, the ontol- of messages exchange during the time-window ΔT{Bx ,SPi } in
ogy responsible for data context characterization is presented, which bus Bx is located at stop point SPi .
which is essential for the knowledge extraction in order to
improve the overall system.   

  
 

A. Inter-networking Architecture

 
 !"#$%
In Figure 1 the inter-networking architecture is depicted. All
$% ) 
 !
data collection in the architecture occurs in the communication 
  !"#

between the Buses and the Stop Points composing the public  
  
#$ !     % $ 
transportation system. The proposed architecture is developed !
&$ 
 
considering that each stop point is equipped with a wireless  
   '

access point, in order to perform data exchange with the buses.    
  
  
 
  ($  !
 
  
 !    !
   "     '
 "

 
 !"#&%




)    
&%

  !"#
 

 
 

  


 
Fig. 2. Sequence diagram for the exchange of messages between a given bus
Bx and a given stop point SPi , during the time-window ΔT{Bx ,SPi } .
 
!
"
   
 
As seen in Figure 2, the entire process that occurs in
the time-window ΔT{Bx ,SPi } requires the establishment of a
 wireless communication channel between a given bus Bx and
 a given stop point SPi . Regarding wireless communication

 
      security, this work relies on standard 802.11 technologies, such

as WPA, WPA2 and the use of RADIUS for authentication. As
  
soon as the wireless secure channel is established, the buses
can access the application running at infrastructure available at
Fig. 1. Proposed inter-networking architecture.
stop points, in order to execute the proposed communication
5 IBM - http://www-03.ibm.com/press/us/en/pressrelease/40140.wss (ac- protocol depicted in Figure 2.
cessed on Sept. 13, 2015) So, in order to have stop points compliant with the proposed

873
inter-networking architecture, the main requirement is related the information collected at bus Bx , and en-queued it for
to their connectivity to the Public Transport Data Service delivery to the central repository. SPi is responsible for
available in a cloud-based infrastructure. The amount of traffic assuring that such information are correctly delivered at the
expected to be exchanged in between the stop points and the central repository at the cloud-based Public Transport Data
Public Transport Data Service is proportional to the number Service.
of lines and their respective frequency along the day. Conse- Besides information related to the buses and the passengers,
quently, a single stop point located in a street, serving a small the architecture can also collect information regarding the
number of lines, should require a simpler inter-networking current conditions of the buses, such as, available fuel level
infrastructure, when compared to bigger terminals (central and engine conditions. Consequently, the size of the message
stations), where several lines are offered. ONSITE COLLECTED INFORMATION sent from a given
Defining the sensing capabilities available at buses is out of bus Bx at a given stop point SPi is not fixed, varying
the scope of this work. But, in general, this work assumes that according to the amount of information available from the
a given bus Bx can easily provide timestamps, such as, the sensors installed at the bus. The standard technology adopted
moment when Bx arrives and departs from a given stop point for such data transfer is REST/JSON6 [7], in which objects
SPi . Based on such timestamps (heart beats), the inference of according to the ontology (Section III-B) can be pushed
the required time slot in between two stops [SPi−1 , SPi ], for a from the buses towards the data repository located in the
certain line L is possible, as the inference of the average speed cloud-based Public Transport Data Service, relying in the
in such section [SPi−1 , SPi ] is possible. As seen in Figure 2, communication infrastructure available at stop points.
after established the wireless communication between bus and As seen in Figure 1, the proposed Public Transport Data
stop point, the first NOTIFICATION message, with a flag set Service is based on the cloud infrastructure paradigm, which
to 0, is used to define the instant when the bus arrived at inherently brings flexibility and scalability to the service,
the stop. Conversely, before departing, the bus sends another based on the concepts of elasticity and agility offered by such
NOTIFICATION message, with a flag set to 1, to define the paradigm [8]. Basically, such concepts support the service ad-
instant when the bus departs from the stop point, followed justment according to the variable load that can be experienced
by the wireless disconnection. By consistence and security along the days or different seasons along the year.
reasons, both moments are acknowledged by the stop point, There is no restriction regarding the model adopted for the
and the timestamps considered in both moments are provided Public Transport Data Service, i.e., being possible to develop
by the stop points. such service in private cloud infrastructures, public cloud
It is important to mention that such infrastructure is capable infrastructures or any other cloud-based solution available, that
of providing bi-directional communication, allowing not only can offer the needed resources to support the components of
the push of sensed information from the buses towards the the architecture, which mainly are:
central repository, but also allowing the pull of information of • Elastic Web Application: the front-end interface to pro-
interest from the central repository to the buses. Based on the vide the interactions with the stop points and buses. The
time-window ΔT{Bx ,SPi } , whenever bus Bx arrives at stop elastic requirements of such web application are in order
point SPi , the architecture specifies that bus Bx first requests to adjust the service according to the current demand.
information from the central repository, such as, information This feature offers the pay-as-you-go paradigm, which
regarding the next stop point SPi+1 . As seen in Figure 2, significantly contributes to the overall costs control to
such request/response process uses two messages, ONBOARD deploy such system;
INFORMATION REQUEST and ONBOARD INFORMATION • Ontology Database: the central repository that stores
RESPONSE. Such process can proactively operate, relying the objects transported using REST/JSON containing the
on information already available at the stop point SPi , or sensed data pushed from the buses. The storage of such
in a reactive basis, in which the ONBOARD INFORMATION objects is according to the ontology, and such database
REQUEST message triggers a query at the Public Transport is key for semantic queries in order to extract knowledge
Data Service. from the collected data;
Note that during the process described so far, passengers are • Inferring/Mining Engine: composed of a (also elastic and
embarking and disembarking at stop point SPi . There are dif- agile) set of servers running inferring/mining applica-
ferent technological solutions for such passengers accounting, tions. Such engine generates the data pushed back to
including image recognition mechanisms [6], or simpler solu- the buses and stop points, in order to keep the public
tions, such as turnstiles at the doors, or bus weight checking transportation system running. Such engine is also re-
to provide estimations regarding the numbers of passengers in sponsible for providing data to the public transportation
the buses. Whenever such embark and disembark is finished, system designers, detailing to them the overall system
i.e., the doors are closed, the module at the bus Bx gen- behavior, and also assisting them defining the required
erates the message ONSITE COLLECTED INFORMATION improvements to the system. In the IoT context, such
and delivers it to the stop point SPi , which replies with
the respective ONSITE COLLECTED INFORMATION ACK 6 JSON-LD also may be used, which also express the semantic meaning of
message. Such acknowledgment indicates that SPi received data - http://www.w3.org/TR/json-ld (accessed on October 29, 2015)

874
servers are also responsible for the data sharing among the reverse is Part of. A Line is exploited by a Bus and the
other services in a widely-integrated-scenario. reverse is used by. A Bus Journey is Composed of classes Bus
The adoption of such cloud-based infrastructure is essential and Line and the reverse is part of.
for moving from the M2M scenario to the IoT scenario, The intention is to share all data between any entity pre-
where the information collected from the public transportation sented in Figure 1 using the Semantic Web standards, which
system can be available to users through the Internet, as allows inferences and queries that relate to data that are not
considered in the architecture through the Consume API, explicitly linked. Queries about the data may be answered such
from which, for example, Online Users (probably passengers) as “what is the average time to go from stop point SPi to
can obtain information of interest from the Public Transport SPj ?”, “what are the best periods (considering the amount
Data Repository. Moreover, such information can be shared of time spent) to go for the following bus journey?”, “is
among other services in order to deliver integrated solutions, there alternative routes to go from where the user is to some
as proposed in the context of IoT and Smart World. other point in the city?”, and so on. As the data shared has
meaning based on an ontology, the vocabulary and definitions
B. Ontology are previously defined, which allows that new apps could be
As aforementioned, the main idea is to create data that developed and the data could be used freely by users.
may be shared by anything (bus stops, users, buses, etc). The The stop point itself is part of the ontology, and can benefit
approach used here is to model all data using an ontology. On- from collecting information regarding the lines that stop on it,
tologies are used to define a context and establish a common for example, in order to display to passengers at this stop point
vocabulary and definition between all parts of a system/group. information regarding the buses that will arrive in the next
The ontology defined here, presented in Figure 3 is still a minutes. Such information can be enriched with the number of
working in progress and it is based on the work of [9]. The passengers inside the buses, providing a hint to the passengers
classes presented in the ontology are: whether they will be able to embark or not, based on the
historical average number of disembarking passengers.
The information provided to the passengers can indicate


whether the bus is delayed or not and, for example, also

indicating to passengers that a given bus is broken. In such
later case, the system could indicate to passengers to which
 


 

specific stop point they should redirect themselves or if it is


 
 


worth to wait at the same stop point for a next bus. As seen,
 
 
the demands over the system can be more complex, and the
  

proposed ontology is designed to support generating answers


to demands like these.
  




IV. E XPERIMENTAL R ESULTS / I MPLEMENTATION



 


 

We turn our attention now to an experimental evaluation of
 
  
 

this proposal. The experiments were run using a prototype

 

considering an emulated bus line at Uberlândia - Brazil,
composed by four stop points in a total segment of 8.5
Fig. 3. Proposed ontology. kilometers (≈5.3 miles) illustrated on Figure 4, where a single
emulated bus traveled in such line 200 times in each direction.
• Network: a class that express a network of buses lines; As the emulated line runs in both directions, we define Line
• Line: a class that is basically a set of stop points that a 1 as the direction from the central station in downtown to the
bus may follow; BR-050 road, and Line 2 as the opposite direction.
• Stop Point: a bus stop, where users can get into or out of Table I details the distance in kilometers from one stop point
the bus; to the other. It is important to mention that, even though the
• Bus: represents a bus; bus line runs in both directions (Line 1 and Line 2), whenever
• Driver: the person that is responsible for a Line; results are presented in this sections, the notation used to
• Bus Journey: the journey of a bus through a Line, which identify the stop points is statically defined, considering as
has information as hour, date, etc; stop point 1 – SP1 – the central station at city downtown (Av.
• Bus Journey Part: represents a segment betweem two Américo Salvador Tangari 11-13); stop point 2 – SP2 – at Av.
Stop Points for a Bus Journey. João Naves de Ávila 1352; stop point 3 – SP3 – at Av. João
The Bus Network class has the relationship Composed of Naves de Ávila 2675; and stop point 4 – SP4 – near BR-050,
the class Line, and the reverse relation is Part of. A Driver at Av. João Naves de Ávila 6970-7000.
has the relation is Reponsible for a Line, and the reverse is Figure 5 illustrates the implementation we produced in our
served by. A Line is Composed of the class Stop Point and prototype. The prototype has the following configuration:

875
Fig. 4. Emulated bus line composed of four stop points. Line 1 runs from Av. Américo Salvador Tangari 11-13 (central station) to Av. João Naves de Ávila
7000 (BR-050 road). Line 2 performs the opposite direction.

TABLE I running Linux Mint 17.1 Rebecca (64 bits) with ker-
D ISTANCE BETWEEN STOP POINTS IN THE EMULATED SCENARIO . nel 3.13.0-37. Its wireless interface chipset features a
Segment [SP1,SP2] [SP2,SP3] [SP3,SP4] RTL8188CE 802.11b/g/n WiFi Adapter.
Distance 2.3 km 1.4 km 4.8 km

Emulated
Bus

• Public Transport Data Service: It was implemented as


a cloud host at Amazon EC2, using an instance type
t2.micro, running Ubuntu 14.04.2 LTS with kernel 3.13.0-
48. The instance has a single Intel Xeon core up to
3.3GHz and 1 GB RAM memory. The central repository
was implemented using MySQL Server version 14.14
Distribution 5.5.44, for debian-linux-gnu x86 64 using SP3
SP1
readline 6.3;
M
ikr
• Stop Points: The four stop points were implemented on oti
kR
Public Transport

top of a single Mikrotik RouterBOARD 800, equipped SP2 ou SP4


Data Service

ter
with four miniPCI slots where four R52N wireless in-
terface cards are installed, three Gigabit Ethernet ports,
two daughter-board connectors, a miniPCI-e slot and a Internet
compact flash slot. Its CPU features a MPC8544 800MHz
Amazon EC2
core, 256MB DDR2 SDRAM on-board memory and data
storage on NAND memory chip. The RouterBOARD is Fig. 5. Details about the prototype implementation.
running MikroTik RouterOS v4 with Level6 license. The
stop points run as MetaRouters on top of RouterOS, Besides the prototype realization, which is part of the con-
running OpenWrt ATTITUDE ADJUSTMENT (bleeding tributions presented in this paper. The experiments that we run
edge, r31206), in which the modules for the Web Server allowed us to analyze the costs for the communication between
Apache 2.2.15-3, PHP5 and Mysql Client: 5.1.53-7 were the emulated bus and stop points. The results presented in
installed. So, the RouterBOARD runs four MetaRouter this section are two fold. The first results elaborate about
instances for the emulated scenario, each MetaRouter the communication cost of the messages pushed from the
assigned to an individual R52N wireless interface; bus towards the Public Transport Data Service and, in the
• Bus: It was implemented using a notebook with Intel(R) sequence, the results present about the RTT (Round Trip Time)
Core(TM) i3-2330M CPU @ 2.20GHz with 8GiB RAM, required to pull information from the central repository to the

876
bus. of embarking and disembarking passengers, fuel and engine
The first analysis presented in Figure 6 provides information information, and others, all defined as attributes of the class
for the RTT required when a bus requests information from Bus of the ontology). Immediately before departing from SPi ,
the central repository using the ONBOARD INFORMATION a message ONSITE COLLECTED INFORMATION is gener-
REQUEST message. Such analyses correspond to the process ated, as described in Figure 2, containing all such information,
depicted in Figure 2, when a certain bus Bx arrives at a and it is delivered to the APi composing the communication
given stop point SPi and recovers, for example, information infrastructure at SPi , responsible for the correct delivery of
regarding the situation at the next stop point SPi+1 to display such message ONSITE COLLECTED INFORMATION to the
to passengers inside the bus. As seen in Figure 6, the average Public Transport Data Service. Consequently, the time to move
RTT varies from ≈ [2.6, 2.8] seconds at the stop points SP3 ONSITE COLLECTED INFORMATION from bus Bx to APi
and SP4, increasing to ≈ [3.2, 3.6] seconds at stop points SP1 under normal conditions (failure-free) is expected to be short,
and SP2. being not evaluated in this paper.
The results presented in Figure 7 show the aver-
age size in bytes of the messages ONSITE COLLECTED
Response delay (ms)

3600 Bus Line 1 INFORMATION generated at the emulated bus in order to be


Bus Line 2
3400 pushed to the Public Transport Data Service of our prototype.
3200
The emulated sensors in our experiments include information
regarding the number of embarking and disembarking pas-
3000 sengers and information according to the proposed ontology.
2800 The obtained values in the experiments are very stable, ≈
132 bytes per message. Basically, there is a single emulated
2600
1 2 3 4 bus, composed of the same amount of sensors in the entire
Stop Points evaluation. In the real world, a different number of sensors is
expected to be present in the buses, leading to more variations
Fig. 6. Average RTT for performing queries from the emulated bus to the in the size of the messages.
Public Transport Data Service.
HTTP Request Content Size (bytes)

In the evaluations, we considered that none information 135


was proactively available at the stop point infrastructure. So, Bus Line 1
the experiments consider the time from which the ONBOARD 134 Bus Line 2

INFORMATION REQUEST messages are issued at the emu- 133


lated bus (notebook), passes through the infrastructure avail-
able at the emulated stop points (MetaRouters running on 132
the Mikrotik RouterBOARD), arrives at the Public Transport 131
Data Service (machine at Amazon EC2) which generates the
ONBOARD INFORMATION RESPONSE messages and push 130
1 2 3 4
them back to the emulated bus.Such time can vary according to Stop Points
several factors, including variable network traffic and variable
load on the servers located at the Public Transport Data Fig. 7. Average size of messages containing sensed data collected from the
Service. We consider that the network traffic variation aspect emulated bus.
was covered in these experiments, since the communications
with the server machine at Amazon EC2 is through public Even considering the same bus, and the same amount of
Internet. At the same time, variations on the servers’ load were sensors, as seen in Figure 7, there are fluctuations in the
not considered in the experiments, but based on the elastic size of the messages. This occurs due to the REST/JSON
characteristic of cloud-based infrastructures, we consider that nature of objects serialization in the messages payload, when
the service work-load can be adapted to better accommodate factors such as variable number of characters in the objects
such fluctuations. It is also possible to apply prediction models result in variations in the size of the messages. In essence,
in order to balance the load on the servers, for example, the collected ONSITE COLLECTED INFORMATION mes-
considering the proposal in [10]. sages include key information that can help improving the
Therefore, we consider the obtained average RTT values overall quality of the public transportation system, and the
as representative to the proposed ONBOARD INFORMATION current average size of ≈ 132 bytes can significantly be
REQUEST/RESPONSE communication process, and can fea- extended, considering an upper-bound around 1500 bytes, in
sible occur in the time-window ΔT{Bx ,SPi } , expected to be order to avoid the fragmentation of ONSITE COLLECTED
in the order of tens of seconds. It is also important to mention INFORMATION messages, what can raise the amount of time
that, according to the proposal, while at a stop point SPi , required to push such information towards the Public Transport
bus Bx is collecting information from its sensors (number Data Repository.

877
V. C ONCLUSIONS AND F UTURE W ORK [4] U. N. D. of Economic and S. A. S. Division, International standard in-
dustrial classification of all economic activities (ISIC), rev. 4 (Statistical
This paper detailed the proposed inter-networking architec- papers). UN; Rev. ed. edition, 2009.
ture and the work-in-progress related to the proposed ontology, [5] J. Höller, V. Tsiatsis, C. Mulligan, S. Karnouskos, S. Avesand, and
D. Boyle, From Machine-to-Machine to the Internet of Things: Intro-
designed to perform ubiquitous data collection from pub- duction to a New Age of Intelligence. Elsevier, Apr. 2014.
lic transportation systems and answering semantic questions. [6] N. Bernini, L. Bombini, M. Buzzoni, P. Cerri, and P. Grisleri, “An
Specifically in this paper, the focus of data modeling is on embedded system for counting passengers in public transportation vehi-
cles,” in Mechatronic and Embedded Systems and Applications (MESA),
bus-based transportation systems. 2014 IEEE/ASME 10th International Conference on. IEEE, 2014, pp.
The results presented in this paper, regarding the RTT 1–6.
and costs for pushing sensed information towards the Public [7] M. Kazuaki, “Performance evaluation of object serialization libraries in
xml, json and binary formats. 2nd int,” in Conf. on Digital Informa-
Transportation Data Service demonstrate the feasibility of tion and Communication Technology and its Applications (DICTAP),
performing the proposed ubiquitous data collection, which we Bangkok, 2012, pp. 177–182.
consider essential for applying the required improvements in [8] C. Esteve, R. Pasquini, F. Verdi, and M. Magalhaes, “Novas arquiteturas
de data center para cloud computing,” in Book Chapter published as a
the overall public transportation network. Mini Course of the 28th Brazilian Symposium on Computer Networks
The proposed architecture can work starting with a period of and Distributed Systems (SBRC 2010), Organized by: CA Kamienski;
data sensing after the deployment of this solution in a city, for LP Gaspary; MP Barcellos. ISBN: 9772177497006. Idiom: Portuguese.
Gramado, RS, Brazil, 2010.
example, in Uberlândia. Based on such initial data sensing, [9] M. Houda, M. Khemaja, K. Oliveira, and M. Abed, “A public trans-
major details from the public transportation system can be portation ontology to support user travel planning.” in RCIS. IEEE,
captured, conducting the first adjustments in the timetables and 2010, pp. 127–136.
[10] M. KOHANA and S. OKAMOTO, “A server selection method for
definition about the required number of buses (frequency) of web-based multiserver systems,” in Advanced Information Networking
the lines. From such moment, the new data collected will help and Applications (AINA), 2015 IEEE 29th International Conference on.
in the stabilization of the system and enrich the information IEEE, 2015, pp. 112–117.
[11] M. Doering, T. Pögel, W.-B. Pöttner, and L. Wolf, “A new mobility trace
provided to the users inside the buses, stop points and from for realistic large-scale simulation of bus-based dtns,” in Proceedings of
the Internet. the 5th ACM workshop on Challenged networks. ACM, 2010, pp.
The prototype also shows that the development of this 71–74.
proposal on top of standard technologies, available on different
low-cost hardware is possible. Thus, the information gathering,
based on the proposed ontology, can significantly help to
improve the overall quality of experience of users of public
transportation systems with reduced technological costs.
As future work, the ontology will be extended and formal-
ized for a wider context. We are also working in large scale
experiments, in which entire transportation networks can be
evaluated with the help of simulation tools [11]. If possible,
we would like to implement our prototype in a certain number
of real bus lines, in order to improve the prototype.
Also, as the amount of data collected rises, and the type of
data as well, new correlations and information about them may
be defined, which will open to new and diverse information
services offered to the population.
ACKNOWLEDGEMENTS
The authors would like to thank CAPES, CNPq and
FAPEMIG for supporting this work.
R EFERENCES
[1] C. Carvalho, “Elasticidade-renda dos gastos das famı́lias metropolitanas
brasileiras com transporte urbano e aquisição de veı́culos privados,”
Instituto de Pesquisa Econômica Aplicada - IPEA, Discussion Papers
1947, 2014. [Online]. Available: http://EconPapers.repec.org/RePEc:ipe:
ipetds:1947
[2] C. Carvalho and R. Pereira, “Gastos das famı́lias brasileiras com
transporte urbano público e privado no brasil: Uma análise da
pof 2003 e 2009,” Instituto de Pesquisa Econômica Aplicada -
IPEA, Discussion Papers 1803, 2012. [Online]. Available: http:
//EconPapers.repec.org/RePEc:ipe:ipetds:1803
[3] IPEA, “A mobilidade urbana no brasil, infraestrutura social e urbana no
brasil subsı́dios para uma agenda de pesquisa e formulação de polı́ticas
públicas,” Instituto de Pesquisa Econômica Aplicada - IPEA, Bulletin 94,
2011.

878

Anda mungkin juga menyukai