Abstract
Traffic Congestion is a very serious problem which is
becoming ever worse as the growth in the number of cars on the road
significantly outpaces the provision of road capacity. This paper presents a
novel Traffic Management System (TMS) for Wireless Vehicular Networks that
combats this problem by seeking to optimize the usage of the existing road
capacity. It also outlines an architecture which includes a novel server-side
decision making module, that enables the dissemination of instructions to
vehicles; if followed these result in optimal road usage.
Keywords:
Traffic Management, Road Vehicles, Driver Instruction, Wireless Networks
1 Introduction
Wireless Access in Vehicular Environments (WAVE) has
been an active research area for some time.WAVE is still in the standardization
phase and several groups are working to that end, such as the Car
2 Car Communication Consortium, the Dedicated Short
Range Communications working group and the IEEE 802.11p task group. Regardless
of the outcome of this process the potential for the provision of new
applications in the vehicular realm is evident, be they Internet based,
convenience, safety or traffic management applications. Much of the early
research into viable use-cases for WAVE has focused on providing a safety
benefit. The Vehicle Safety Communications Consortium has created a long list of
distinct use-cases ranked according to their benefits. The majority but not all
of these are safety related. At present many projects worldwide are developing
safety applications based on these or similar scenarios.
More recently academics have begun to branch out into
different research directions involving WAVE. Some applications under
development include: MobEyes: a proposal to use vehicles as sensors in a mobile
surveillance platform for urban monitoring , FleaNet: a virtual ”flea market”
allowing drivers and roadside shops to advertise their demands/offers and free parking space discovery. One use case
for WAVE which has enormous potential is Traffic Management. Vehicular traffic
is one of the great ills of modern society; in urban areas commuters can spend
a significant percentage of their life stuck in traffic. It has been estimated
by the Texas Traffic Institute that traffic congestion will
cost the US over $90bn per year by 2009 and the UK
Treasury put the cost to its country’s economy at £20bn (US$38bn) for
2006. These are huge monetary costs, based on lost productivity and wasted
fuel but there is also the environmental cost to
consider. Disturbingly the situation is set to worsen as the number of vehicles
on the road outpaces growth in road capacity worldwide. From 1982 to 2002, the
number of vehicles in the US
grew by 36% and vehicle miles travelled by 72% while road capacity increased by
less than 5%. Between 1990 and 2004 the number of cars in the 25 EU member
states rose by over 40% and continues to rise, from 1990-1998 total length of
motorways in the EU grew by 28% but has remained roughly stagnant since then. This
paper proposes TraffCon: a novel
Traffic Management System (TMS) for WAVE (figure 1).
The main aim of TraffCon is to improve the overall
Quality of Driving Experience (QDE). To achieve this objective the overriding
emphasis of such a system is to reduce traffic congestion/increase traffic flow,
by maximising the usage of the available infrastructure. TraffCon’s benefits
are varied: social, economic and environmental i.e. shorter journey times,
financial savings, increased productivity and a reduction in vehicle emissions.
The rest of this paper presents the proposed TMS.
2 RelatedWorks
There are many research groups exploring use cases
for WAVE which improve QDE by influencing traffic conditions. These can be
loosely divided into three main categories Traffic Information/Advisory Systems
(TIS), Autonomous Vehicle Systems and Traffic Management Systems.
2.1 Traffic
Information/Advisory Systems (TIS)
A number of TISs have been developed i.e. systems
which gather traffic data and disseminate traffic information to users, so they
can make better informed decisions regarding their route. Examples of this
include TrafficView: a device which gives drivers an extended horizon i.e. a
real time view of road traffic far beyond what they can actually see,
StreetSmart: a system which identifies and disseminates traffic patterns to
users and SOTIS: a system which distributes up-to-date travel and
traffic information pertinent to a vehicles locale. While
these systems do keep drivers better informed about traffic conditions, there
is no telling how the driver will interpret the information given. Consequently
there is no guarantee such systems lead to more beneficial or optimal route
decisions. Much work has been also done exclusively in the area of Data
Harvesting and Information Dissemination schemes for WAVE
2.2 Autonomous
Vehicle Systems
Autonomous vehicle systems can provide traffic
control solutions by fully automating vehicles and thereby removing user
responsibility for driving. There has been and continues to be a wealth of
research in this area, the most celebrated of which feeds into the DARPA Grand
Challenge. Some notable recent work includes a vehicle capable of navigating
complex environments using artificial vision, a perception and planning
architecture for autonomous vehicles and a system capable of avoiding complex
obstacle filled environments to complete a journey described by a simple set of
waypoints. However at present such solutions are prohibitively expensive for
large scale deployment and must also overcome the challenge of user resistance
to automation.
2.3 Traffic
Management Systems (TMS)
Systems which actively control aspects of the traffic
network in order to force member nodes into a behaviour which has some benefit
to the system as a whole can be classified as TMS’s. Current work in the area
includes adaptive traffic lights for improved traffic co-ordination at
intersections and train - vehicle communications to manage their interactions
at road and rail intersections.
3 Traffic
Management System (TMS)
There are enormous challenges in developing a fully
functional large scale TMS, i.e. for a large urban area. Even for a modest
sized urban area such as the town of Cambridge (UK)
population 100,000, there are 183,850 vehicles passing through it in the 12
hours from 07:00 to 19:00 on a typical day. When
a large metropolitan area is considered it is clear
that harvesting traffic data from vehicles will yield vast volumes of data.
Storing this data may prove problematic not to mention processing it in
realtime and disseminating control messages. However the first step in the
development is to determine - What can potentially be changed/controlled in
order to alter/manage traffic conditions?
² A vehicles route - Vehicles may be directed
to follow a specific path en route to their destination.
² A vehicles lane - Vehicles may be directed
to change their lane e.g. bus lanes could potentially be used to increase
capacity provided buses are not delayed.
² Vehicle speed - Vehicles may be instructed
to adjust their speed.
² Traffic light interval times - The green
light times and ratios (i.e. favour one road over another at a junction) may be
adjusted. A simple indicator of traffic congestion is the ratio of the number
of vehicles on the road to road capacity. Re-routing and adjusting lane
positions of vehicles allows the road capacity to be maximized thereby reducing
congestion and increasing the flow of traffic. Spaces between vehicles occupy
road capacity in the same way vehicles do. If vehicle speed can be controlled
to minimize spaces between vehicles, then road capacity can be further
maximized. When traffic lights are red they disrupt the flow of traffic;
optimizing traffic light operation to make traffic flow as arterially as
possible is clearly beneficial. While manipulating these elements of the traffic
system can improve traffic flow other factors must
not be neglected in the quest for speed. The stress
of sitting in traffic should not be replaced by an irritating or overly
invasive interface directing drivers in the cockpit, the system should not
force drivers to drive in a manner which is erratic or uncomfortable and it
should in no way endanger the safety of the driver. In short for any TMS safety
is paramount, reducing journey time is vital, but comfort is important
too. In the future there may be other infrastructural
elements a TMS could interact with e.g. future roads may be designed so that
their layout can be altered (i.e. painted white lines are replaced by some form
of electronic display which can be modified).
4 TraffCon:
Intelligent Traffic Control Solution
This paper focuses solely on managing vehicle routes.
It is assumed that vehicles have a GPS receiver connected to a computing device
with wireless connectivity.
4.1 System
Architecture
TraffCon has a client server architecture. Vehicles /
client nodes communicate with server nodes responsible for traffic management.
The systems functional blocks are divided between client and sever as shown in
figure 2. Server side decision making means instructions are disseminated to
clients. This architecture is what differentiates this TMS from traditional
Traffic Information/Advisory Systems (TIS) where information is disseminated to
the clients and drivers are responsible for decision making as seen in figure
3.
4.2 System
Functional Blocks
The system is comprised of four main functional
blocks:
1. Data Harvesting - all nodes in the system gather
useful traffic data.
2. Data Processing - the data is filtered, aggregated
and refined to generate precise information regarding the state of the traffic
network.
3. Decision Making - the traffic network information
is used in a decision making process which generates a route instruction which
if followed has a benefit over the other route choices available e.g. improved
traffic flow, a reduction in fuel consumption
4. Instruction Consumption - the instruction is
consumed i.e. it is followed or ignored.
4.2.1 Data
Harvesting
All vehicles in the TraffCon system will gather data
regarding the state of the road network. In order to understand the data
collected, it is necessary to define two components which can be used to
describe the road network-
² Junction: Point where two or more roads meet
² Link: Section of road between two junctions.
For two junctions J and K joined by a single section
of road, there are two links connecting them i.e. the link JK which allows
traffic to travel from J to K and the link KJ which allows traffic to flow from
K to J. It is assumed all vehicles carry map data for the area in which they
travel. At the most basic level this data is simply the GPS co-ordinates of all
junctions in the area. Given that all nodes know their own location (from their
GPS receiver) and the locations of all junctions in the area, then a node can
identify when it has reached a junction. Initially two pieces of data are
obtained from a vehicle before it begins its journey i.e. starting location and
destination location. While in transit a vehicle constantly checks whether it
is at a junction; if a junction is reached a timestamp is set. Whenever a
vehicle has traversed a link the time taken to do so (i.e. the link time), is
calculated. This link time, the time the vehicle entered the link (reached the
first junction) and the link ID (two GPS co-ordinates i.e. longitude-latitude
pairs; J and K for the junctions at the beginning and end of the link combined
in the form JK) are packaged and sent to the server. Table 1 shows the data format a small sample set of data passed by
clients to a server.
By gathering such simple data the server can generate
a wide range of useful information such as; average link times, average link
speeds, instantaneous per link vehicle density, etc.
4.2.2 Data
Processing
The data set described above is used to generate a
table of average link times in the format shown in table 2 by using window-based averaging.
Given that the server also knows the length
of every link then a table of average link speeds is also generated using speed
= distance/time.
4.2.3 Decision
Making In a TIS the user is responsible for making route decisions. They
are given real-time traffic information and
it is assumed they can make route adjustments that are beneficial to
themselves. As a result the system
is tailored to only benefit individual client nodes. The emphasis is solely on
getting individuals to their
destination as quickly as possible, with no consideration for the effect on the
overall traffic system. For the TMS
with server side decision making, the overall situation is of paramount
importance and vehicles are given
route instructions designed to benefit both the individual and the overall
system.
Genetic Algorithms are used as they are a suitable
approach for solving such a combinatorial optimization problem. The fitness
function eq. 1 is proposed to make route decisions which minimize journey time
and fuel consumption. Parameters for overall and individual benefit are used.
F(y) = w1=J(y) + w2=I(y)
+ w3=E(y) + w4=D(y) + w5S(y) (1)
Where; J(y) is average journey time in the system,
I(y) is individual node’s journey time, E(y) is average fuel consumption in the
system, D(y) is individual node’s fuel consumption, S(y) is solution fairness
(designed to keep a balance between the benefit to the individual and to the
system as a whole
such that no individual is overly rewarded/penalised)
and wi are weighting factors. The information made available in the data
processing stage is pulled as required to evaluate parameters. For Example trip
times can be estimated by summing the average link times along a route. It is possible
to enhance this function at a later date by considering additional parameters
e.g. speed and/or jitter.
4.2.4
Instruction Consumption
An interface of some kind is required in TraffCon
enabled vehicles to present instructions to the driver for consumption. Options
include an audio solution or a visual solution such as Head-Up Display (HUD). Regardless
of interface drivers should receive pertinent instructions in a timely fashion.
5 Testing
In order to evaluate the proposed solution the Java
in Simulation Time / Scalable Wireless Ad hoc Network Simulator (JiST/SWANS) is
used in conjunction with the Street Random Waypoint (STRAW) - vehicular mobility
model for network simulations. The vehicular mobility model employed by the
simulator uses real world road maps as seen in figure 5. This simulator setup
allows the wireless network and vehicular mobility aspects of the system to be
simulated simultaneously. The existing model was enhanced with data harvesting
and data processing modules which follow the descriptions in sections 4.2.1 and
4.2.2 respectively. No decision making module has been deployed as yet. It is
planned to model the consumption of instructions with varying percentages of
obedience. A potential simulation configuration is shown in figure 4; client
nodes move in a field according to STRAW Street Mobility and run the TraffCon
Client Application, TCP is used at the transport layer, IPv4 at the network (in
conjunction with Greedy Perimeter Stateless Routing (GPSR)) and 802.11b at the
MAC, in this setup. There is a single sever node which runs the TraffCon Server
Application, and is static rather than mobile but is otherwise identically
configured. In order to evaluate system performance the behaviour of TraffCon
will be contrasted against two other benchmark systems, across a range of
parameters: average node - speed, journey time, fuel economy, jitter etc.
The testbed will include:
1. A model where all vehicles attempt to take the
shortest route to their destination (in spatial terms).
2. A model where vehicles run a TIS which instructs
them to take the quickest route, by using, Dijkstra’s Algorithm to make route
decisions based on average link time information supplied by
the server.
3. A model where vehicles run the proposed
TraffCon-based system. The first system models the real world behaviour of
vehicles with no navagational aids. The second examines what happens when
vehicles are routed optimally but in a greedy fashion i.e. with no regard for
the effect on other vehicles. The third will show the result of routing
vehicles optimally with concern for the overall system. In both the second and
third cases the effect of varying the penetration rate of the technology will
also be examined.
6 Conclusion
Traffic congestion is already a major problem
worldwide and it is becoming more and more serious because the number of cars
on the road is increasing at a higher rate than road capacity. In this context
this paper has introduced TraffCon a novel traffic management system for WAVE,
which aims to optimize the usage of existing road capacity. The system
architecture has been fully outlined and three main functional blocks Data
Harvesting, Data Processing and Decision Making have also
been described in the detail. Further enhancements are envisaged to include the
addition of a feedback loop to attribute congestion charges/credits to drivers
based on whether they disobey/comply with instructions received. Such a
penalty/reward paradigm adds greater likelihood of compliance and brings the
system closer to true control. Vehicle lane and traffic signal control may also
be added. These enhancements would help to further maximize road capacity.
REFERENCES:
Ø
“Intelligent Transportation Systems, U.S.
Department of Transportation.” http://www.its.dot.gov.
Posts
