FRIEND Cyber-Physical System for Traffic Flow

FRIEND:
A Cyber-Physical System for Traffic Flow
Related Information Aggregation and
Dissemination

Samy El-Tawab
Advisor: Professor S. Olariu
PhD Defense
Intelligent Networking and Systems (iNetS) Research Group
Department of Computer Science
Old Dominion University
PhD Defense Samy El-Tawab July 27th , 2012

Outline
 Introduction
 Motivation and background
 Objectives and goals
 System infrastructure
 Physical components

 Reasoning about traffic flow parameters

 Communication protocol in FRIEND

 Decision making in FRIEND

 Summary
 Future research ideas

Driving on highway

If you are driving on highway
What would you need?

Image reference: www.driversedguru.com


Introduction
• Vehicular Ad-Hoc Network (VANET)
is a type of Mobile ad-hoc network
(MANET) that allows communications
between nearby vehicles and between
vehicles and roadside infrastructure

• Intelligent Transportation Systems
(ITS) are advanced appli-cations which
aim to provide innovative services
relating to different modes of transport
and traffic management and make safer,
more coordinated, and „smarter‟ use of
transport networks http://media.nowpublic.net

More about Vehicular Networks…
• Main characteristics
– uses vehicles as network nodes
– road side units as fixed nodes
– nodes move relative to each other but within the
constraints of the road infrastructure
– limited communication range
– high mobility of nodes


VANET – a closer look
• Vehicle-to-Vehicle (V2V)

• Vehicle-to-Infrastructure (V2I)

• Infrastructure -to-Infrastructure (I2I)


And more about ITS…
• Responsibilities:
– basic management systems :
• car navigation
• traffic signal control systems
• automatic number plate recognition
• speed cameras
– to more advanced applications:
• monitor applications: security CCTV systems
• parking guidance and information systems
• weather information
• bridge de-icing systems


However…
• Although all these sensing technologies:
– inductive loop detectors
– magnetic sensors – passive and active
– ultrasound sensors
– infrared sensors
– microwave sensors
– laser sensors
– video image processors


Even VANET
• A typical VANET system for reporting traffic conditions
consists of vehicles exchanging information about their
position and speed with each other

• The vehicles then use this information to determine where
traffic slowdowns are occurring and report that information to
other vehicles
– NOTICE: an architecture for the notification of traffic
incidents


VANET applications

• Traffic monitoring: to monitor the highway, to give
information about the flow, speed and density on road

• Incident detection: to detect and notify drivers about incidents
on highway

• Weather alerts: ice, foggy, heavy rain and tornado watch

• Emergency situations: closed road, maintenance and planned
evacuations


Our classification of VANET applications

Real-Time Traffic Monitoring

Incident Detection Traffic Information
System

Weather Alert System Backup Warning
System

Data Collection on
Highways

Temperature
Monitoring on
Highways

What we need?
• Integrating resources and capabilities at the nexus between
the cyber and physical worlds, (a cyber-physical system for
traffic flow-related information aggregation and
dissemination) FRIEND will contribute to aggregating traffic
flow data collected by the huge fleet of vehicles on our roads
into a comprehensive, near real-time synopsis of traffic flow
conditions

• We anticipate providing the drivers with a meaningful, color-
coded, at-a-glance view of flow conditions ahead, alerting
them to traffic events


Motivation

• FRIEND explores the integration of wireless
networking with lightweight roadside infrastructure
into a cyber-physical system (CPS) that enables

– privacy-aware detection of traffic-related events

– the dissemination to the driving public of such aggregated
information both in the form of a color-coded traffic status
report and traffic advisories in the case of serious incidents


Objectives and goals – in brief

• To collect traffic data about the traffic flow

• To aggregate the collected data in a way that allows
to detect and/or to anticipate traffic-related events

• To disseminate relevant traffic related information to
the driving public


Objectives and goals using V2I or I2I

• Traffic data collection
• Traffic status dissemination V2I
• Traffic advisories dissemination

• Acquiring coarse-grain incident location information
• Acquiring fine-grain incident location information I2I
• Acquiring fine-grain information about backup
dynamics


Problem definition

By using already existing infrastructure: to aggregating
traffic flow data collected by the huge fleet of vehicles
on our roads into a comprehensive, near real-time
synopsis of traffic flow conditions and provide the
drivers with a meaningful, color-coded, at-a glance view
of flow conditions ahead, alerting them to any traffic
event


FRIEND - physical components
smart cat‟s eyes (SCE)
• The cat‟s eye nodes are deployed uniformly along the road on
both sides as lane separators

• The intention is for smart cat‟s eyes (SCEs) to replace, in the
near future, the ubiquitous cat‟s eyes

*Photo Credit :http://www.catseyeroadstuds.com

smart cat‟s eyes (SCE) - components
• Architecture: Each SCE is a compact, self-contained package.
It contains several types of sensors (including magnetometers),
a radio transmitter, an RFID tag, a micro-controller, a solar
panel and a lithium battery

• Power consumption
• Communication technology: SCE features a narrowband
frequency-shift keying (FSK) data transceiver as well as one
of many possible types of RFID tags


smart cat‟s eyes (SCE) - features
• suitable for edge line of road and pavement

• can work more than three years which promotes energy efficiency and
environmental friendliness

• load-bearing more than 20 ton as two reinforced veins are designed on the
top

• edges to strengthen compression resistance and protect the solar panel
against compression

• waterproof and unbreakable: the solar panel, electronics and optics are
fitted inside


roadside units (RSU)
• RSU: deployed at regular intervals - consists of
– GPS
– radio transceiver
– a laptop-class embedded computing device
– on-board battery packs charged by solar panels

• Role:
– To collect and aggregate traffic-related information from
the passing cars as well as by interchanging information, on
an intermittent basis with adjacent RSUs


roadside units (RSU) – does it exist?
• Examples from interstate 64 highway


FRIEND - the vehicular model

• Event Data Recorder (EDR)
• GPS receiver
• Wireless transceiver
• Digital built-in map
• Radar
• Smart wheels
– Temperature sensor
– Electronic stability control system
– RFID reader


the event data recorder - EDR
• The EDR record transactions that occurs in the previous area
• These transactions contains: time, location, max speed, min
speed, lane changing


EDR - transactions


FRIEND – traffic terms

• Historical data

• Speed, density and flow rate

• Headway distance, safe headway distance


FRIEND – evaluating


FRIEND – evaluating - conclude

• Evaluating the probability of large headway distances in co-
directional traffic

– Question: given that m cars are deployed uniformly at random in a
single lane of traffic of one kilometer and given that dependable radio
communications between cars require a maximum inter-car distance of
200 meters2 what is the probability that there is end-to-end radio
connectivity between the m cars?

– Answer: the number of cars per kilometer must be at least 16 in order
to have a better than even chance for connectivity, it takes about 23
cars per kilometer for end-to-end connectivity to be present with 90%
probability


FRIEND – evaluating - clarify

• If there were 12 co-directional cars in the window, the
probability of no end-to-end connectivity between them would
be about 86%.

• The probability decreases with the number of co-directional
lanes of traffic in each direction


FRIEND – more evaluating

 Evaluating the expected size of a cluster

• Where m cars and n the number of inter-car spaces and d
corresponds to the maximum effective transmission range


FRIEND – average headway distance

• The following are the steps performed by the RSU to calculate
the average headway distance
– RSUi (in time period [t0, t1]) receives number of records from different
vehicles, its record includes time, location, speed and lane
– RSUi calculate the location of each vehicle at time T within same lane
– RSUi sorts the records and calculate the headway distance between each
vehicle
– RSUi update headway buffer with headway distances recorded
– RSUi compare the headway in the buffer with any received headway data
from vehicles
– the recorded data in the buffer can give us an indication for the traffic
density on the highway at the RSUi


FRIEND – communication protocols
adjacent RSUs
• To detect initial stages of congestion or when an incident
having occurred in the segment between them triggers changes
in the traffic flow

• To gain the fine-grain determination of the location of the
accident

• To support the propagation of the color-coded traffic status
reports to vehicles along the roadway


RSU communication with vehicles


communication from SCEs to RSUs
• We use simple narrowband FSK radio data transmitters that
turn on within milliseconds, and draw only 10-20mA

• Adjacent-channel interference and jamming are very real
problems, but can be mitigated by using a frequency-agile
narrow-band system

• Since this communication does not require a high data rate, we
choose to use narrow-band FSK data transceivers in SCEs


communication between adjacent SCEs


communication from vehicles to SCEs
• FRIEND assumes the use RFID technology as the
communication medium between the smart wheels and SCEs
– the RFID reader in the smart wheels allows the vehicle to inform the
SCEs about speed, stability loss due to road conditions (if any) and
ambient temperature
– the SCEs collect data sent from vehicles every t, where t depends on
highway conditions
– the RFID reader in the smart wheels transmits an object identity using
electromagnetic waves in the SCE, an RFID tag stores its ID in
memory
– the RFID reader which is installed in the vehicle wheels emits RF
radio waves eliciting a signal back from the tag. We use RFID with
radio range (up to approximately 3m)


FRIEND – making traffic-related
decisions – color-coded states
• Level of Service (LoS) is a measure used in ITS by traffic
engineers to assess the effectiveness of various elements of
transportation infrastructure
– A = Free flow
– B = Reasonably free flow
– C = Stable flow
– D = Approaching unstable flow
– E = Unstable flow
– F = Forced or breakdown flow


decisions – color-coded states - transitions
• In order to avoid spurious transitions between colors, FRIEND
has a built-in “laziness” that records traffic flow trends without
necessarily taking immediate action

• Example: the reported status is yellow if the internal Markov
chain is in any of the three yellow states


decisions – color-coded states – mapping
• FRIEND employs to effect state transitions in the internal
Markov chain are
– the average headway distance (AHD)
– maximum speed aggregated (Speed)
– historical data collected over a longer time of monitoring data at the
same locale


decisions – incident classification

Incidents

Blocking Moving Non-
Incidents Incidents blocking
Incidents


decisions – incident expected flow 1/3


decisions – incident detection algorithm
• Task 0: RSU initialization: Initially, we assume that RSUi
just started to collect data

• Task 1: Incident detection: RSUi is notified of an incident or
RSUi notices change of speed or density of RSU-RSU[i,j]

– A notification of lane changing in the same location in the previous
RSU-RSU area in a short time, identifies the possibility of an incident

– Threshold Thi can be determined from historical data, the higher the
threshold the more time needed to detect an incident and the less
chance to generate alarms


decisions – incident detection algorithm
– Task 1-1: Identifying RSU-RSU: Determining which RSU-RSU[i,j]
area where incident occurs ”Global view”

– Task 1-2 : Identifying segment and location: Identify segment with
incident; vehicles that changed lanes in the last segment report lane
change Lc and location of lane change

– Task 1-3: Classifying the incident

• Task 2: Information dissemination


decisions – incident information dissemination
• Different types of events or incidents requires different levels
of propagation depending on how critical the incident and how
long it stays

• Drivers would like to receive information that affects their
decision rather than just notification about incidents that will
be solved by the time they reach this point on the highway

– GPS with life traffic information can give warning messages about
incident that far away from other vehicles
– Virginia 511 offered by Virginia Department of Transportation
(VDOT) is a similar example of a service that disseminate information
on a website or mobile application

decisions – incident information dissemination
• In FRIEND, the more the incident stays, the further the
information will be propagated

• FRIEND compares different densities with the level or
distance of propagation bearing in mind the principle of
locality
• We have two aims for information propagation
– to prevent secondary accidents  Stage I
– notify drivers far away from the accident of an expected
delay by updating there coloring system  Stage II


decisions – stage I
• Focus on the first goal which notifying vehicles with short
distance to an accident
– the RSU is responsible of informing the previous RSU immediately of
the incident in inform the vehicles passing beside it of the incident
– the more time the incident takes to be cleared, the more frequently
previous RSU will be informed of the incident


decisions – stage II
• We obey two rules:
– to track the source of the incident to be able to track the
movement of vehicles after the event is cleared
– To send a long time to live message every T, this message
target far away vehicles in order to be able to take the
decision of keep going or take an exit

• The decision of switching between stages I and II depends on
the average headway distance (AHD), speed of vehicles and
historical data, time and day of the incident


decisions – stage II – track head and tail
• Head of a backup

• Tail of a backup

• Knowing the length of the backup and tracking the Head and
Tail are important information that can be propagated and used
in Stage II to inform approaching vehicles of an incident at a
specific location


decisions – stage II – track head and tail


decisions – stage II – information sent
• The information sent between adjacent RSU(s) is the
following
– Time: the time of last update
– Head location
– Tail location
– Incident clearance flag
– Average speed of arriving vehicles at the RSUt
– Average speed of moving vehicles at the RSUh


FRIEND – evaluation

• The use ONE simulator
– We adopted a two-lane highway similar to Interstate US13
highway in Virginia, USA
– The model assume fixed nodes between the two lanes
which represents SCEs along the highway
– Another fixed nodes every one mile
– Highway length approx. 11miles
– Max speed for vehicles 55 miles/hr
– Model movement : Map based movement


FRIEND – measure

• Our model compares the ratio of messages dropped over all
messages


FRIEND – measure

• we study the idea of having two, three or four SCEs detecting
vehicles at high speed


FRIEND vs. Virginia 511

• VDOT lately latched a system that ( telephone, mobile
application and website)
• Centralized vs. distributed


Concluding remarks

• We built the complete theoretical system FRIEND
– The strongest point of FRIEND is using infra-structure
already exist
– We defined our nodes in details
– We showed the communication protocol between different
nodes
– We calculated mathematically
• the expected headway distance in free-flow traffic in a single lane L
• the probability of large headway distances in co-directional traffic
• the expected cluster size
– We showed the mapping algorithm between the traffic flow
parameters and the 10 Markov chain states


More concluding remarks

• We built the complete theoretical system FRIEND

– We classified the incidents on highways

– We designed an incident detection algorithm

– We described our information dissemination algorithm
with the two stages

– We showed how to track the backup dynamics


Directions for future work

• Enhance the energy efficiency both of data collection and data
dissemination

• Exploit existing (or anticipated) correlation of traffic data to
put RSUs “to sleep” instead of mandating them to continually
collect data

• Perfect an efficient way whereby the vehicles wake up the
RSUs in sparse traffic


More future work …

• Better understand the triggers that signal to FRIEND trends in
the traffic flow that need immediate action to prevent
congestion from building up

• Evaluate the effect of traffic buildup in the case of a serious
incident
– One idea is to merge two backups on the highway in case of different
incidents occurring at the time and impacting the traffic flow
– Another idea to calculate the expected backup length by time


More and more

• Extending the current simulation model for FRIEND by
incorporating more realistic assumptions

• Hands on: Study the SCE(s) by doing couple of experiments:
– Number of vehicles that a SCE can detect on a highway I-81
– Power consumptions after couple of weeks
– Exact cost of one SCE


Publications, presentations, posters and
book chapters
• Samy El-Tawab, and Stephan Olariu: ”Intelligent Road Detection” in The
College of William and Mary‟s 8th Annual Graduate Research Symposium,
Williamsburg, Virginia, March 2009. My paper was awarded a prize for
Excellence in Scholarship

• Samy El-Tawab, Mahmoud Abuelela, and Yan Gongjun: ”Real-Time
Weather Notification System using Intelligent Vehicles and Smart
Sensors”, First International Workshop on Intelligent Vehicular Networks
(InVeNET 2009) Co-Located with IEEE MASS 2009, October 12th, 2009 -
Macau SAR, China

• Yan Gongjun, Danda B. Rawat, and Samy El-Tawab: ” Ticket-based
Reliable Routing in VANET”, First International Workshop on Intelligent
Vehicular Networks (InVeNET 2009) Co-Located with IEEE MASS 2009,
October 12th , 2009 - Macau SAR, China

More publications …

• Book Chapter: Samy El-Tawab, and Yan Gongjun: ”Safety and
Commercial Applications”, Advances in Vehicular Ad-Hoc Networks:
Developments and Challenges. A book edited by Prof. Mohamed K. Watfa
University of Wollongong, UAE

• Book Chapter: Yan Gongjun, Samy El-Tawab, and Danda B. Rawat:
”Reliable Routing Protocols in VANETs”, Advances in Vehicular Ad-Hoc
Networks: Developments and Challenges. A book edited by Prof.
Mohamed K. Watfa University of Wollongong, UAE

• Poster: Samy El-Tawab, and Stephan Olariu:”Monitoring Queue-ends on
highways using Smart Sensors”, 11th Annual Student Research Poster
Session, Christopher Newport University, VA , USA November 2009


And more…

• Samy El-Tawab, and Stephan Olariu: ”FIRMS: A Framework for
Intelligent Road Monitoring System using Smart Sensors” in the
International Journal of Information Sciences and Computer Engineering,
Vol.1 No.2 2010 pages 1-6.

• Samy El-Tawab ”Integrity, vulnerability and security for Vehicular
Networks” in The Doctoral Consortium of the 2010 IEEE International
Conference on Networking, Sensing and Control April 11-13, 2010
Chicago, IL, USA

• Samy El-Tawab, and Stephan Olariu ”A Cyber Physical System for
Highway Applications in Vehicular Networks” in the 9th International
Conference on Mobile Systems, Applications, and Services- PhD Forum”
June 28th-July 1st , 2011,Washington, DC, USA


Even more…

• Samy El-Tawab, Stephan Olariu and Mohammad Almalag ”FRIEND: A
Cyber-physical System for Traffic Flow Related Information aggrEgatioN
and Dissemination” in IEEE VTP 2012 workshop June 25th, 2012 in the
IEEE WoWMoM 2012, San Francisco, CA, USA


Standard VANET application classification

Applications in
VANET

Safety Commercial
Applications Applications

High Priority Low Priority Monitoring and Entertainment
Safety Safety Service Applications
Applications Applications Applications

Lane detection using GPS

• In FRIEND, SCE(s) can play a good role in this case
• We can also use interpolation to estimate the current location
of the car within the road at each sample point ni

FRIEND Cyber-Physical System for Traffic Flow

Recommended

Recommended

More Related Content

Similar to FRIEND Cyber-Physical System for Traffic Flow

Similar to FRIEND Cyber-Physical System for Traffic Flow (20)

Recently uploaded

Recently uploaded (20)

FRIEND Cyber-Physical System for Traffic Flow