AN INTEGRATED ARCHITECTURE FOR INFOMOBILITY

SERVICES

Advantages of Genetic Algorithms in Real-time Route Planning

C. De Castro, G. Leonardi, B. M. Masini and P. Toppan

IEIIT, Italian National Research Council and WiLab, University of Bologna, V.le Risorgimento 2, Bologna, Italy

Keywords:

Intelligent transportation systems, Real-time route planning, Virus-evolutionary genetic algorithms.

Abstract:

In the ﬁeld of Intelligent Transportation Systems, a key role is played by efﬁcient route planning services.

Such systems have been evolving rapidly, but they still have some restricting drawbacks, such as the lack of a

full support of real-time trafﬁc monitoring and the consequent real-time update of the best route suggested.

In this paper, an architecture is proposed for the management of dynamic path planning and limitations of

traditional search algorithms in these kinds of applications discussed. A variant of the proposed approach is

consequently presented, based on the joint use of virus-evolutionary genetic algorithms for real-time route

planning and trafﬁc forecasting.

1 INTRODUCTION

The latest study on global urbanization conducted by

the Population Division of the Department of Eco-

nomic and Social Affairs of the United Nations was

released in 2007 and predicts that, in 2050, nearly

70% of the global population will be living in larger

cities (UN, 2008).

This immense aggregation of people will surely

pose great challenges to the sustainability of modern

lifestyle, and the problem of an efﬁcient management

of mobility stands out as one of the most relevant

ones. As a matter of fact, densely populated cities

imply the concentration (from the country) and distri-

bution (within the city) of massive amounts of people

and resources (EU, 2010).

In addition to the vast economic importance and

consequences of such situation, urban and sub-urban

mobility is a serious challenge also due to the cir-

culation of large amounts of people and goods in a

relatively small area. This poses hazards to life and

health, especially for children, the elderly, and unfa-

miliar visitors, as well as to the environment.

Urban mobility, in fact, accounts for some 30% of en-

ergy consumption and 70% of transport pollution in

Europe, and this problem is magniﬁed by the increas-

ing population concentration in large cities.

In such a scenario, the efﬁcient management of

trafﬁc is a challenge that governments, industries and

researchers are forced to face worldwide. Private trav-

ellers, commercial road users, and the public sector

are continually searching for new and faster travel

routes and methods.

In this context, one of the most important applica-

tions is the support of real-time, meant as the constant

monitoring of trafﬁc and road conditions, and the con-

sequent possible update of the routes previously sug-

gested. As a matter of fact, the best path in a given

situation can vary when trafﬁc conditions vary and

updates should be notiﬁed to the user in real-time.

Nevertheless, up to now, no simple and marketable

product has been proposed for monitoring trafﬁc and

providing real-time information to road users.

Roads efﬁciency can be substantially improved by

the deployment of Intelligent Transportation Systems

(ITS), which exploit Information and Communica-

tions Technologies (ICT) in order to provide trafﬁc

safety and efﬁciency.

ICT can be considered as the foundation for car-

rying out smart navigation, meant as the paradigm

where mobile entities (vehicles and pedestrians) move

wisely through a given environment, exploiting reli-

able and timely information about trafﬁc conditions.

New solutions are gaining interest: several projects

and consortia (ERTICO, 2010; Car2Car, 2010) and

relevant standardization bodies are working on the

development of new standards, so as to deﬁne com-

mon ITS communication architectures to let vehicles,

roadside units, and wireless infrastructures communi-

300

De Castro C., Leonardi G., M. Masini B. and Toppan P..

AN INTEGRATED ARCHITECTURE FOR INFOMOBILITY SERVICES - Advantages of Genetic Algorithms in Real-time Route Planning.

DOI: 10.5220/0003057803000305

In Proceedings of the International Conference on Evolutionary Computation (ICEC-2010), pages 300-305

ISBN: 978-989-8425-31-7

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

cate.

However, only the knowledge and appropriate

processing of actual trafﬁc conditions, as well as their

forecasting, can make the difference in route plan-

ning applications. The problem, thus, is (at least)

twofold: on the one hand, an efﬁcient integrated ar-

chitecture must be designed for the management of

trafﬁc and vehicular mobility; on the other hand, such

architecture must rely on appropriate route planning

algorithms.

As for the architecture, in this continuously evolv-

ing scenario, the Italian PEGASUS project (PEGA-

SUS, 2010) represents one of the ﬁrst initiatives

proposing an infrastructure really able to attain smart

navigation in the short-medium term, on the basis of

actual trafﬁc conditions.

The need of an efﬁcient support for real time, though,

focuses the attention on the second aspect mentioned

above, i.e. the use of algorithms capable of adapting

to the changing environment.

In fact, even though efﬁcient solutions and traditional

algorithms are still being applied in most commercial

systems, the ever growing dimension of the problem

suggests one should reconsider such solutions in the

broader context of intelligent infrastructures and en-

vironments.

As a matter of fact, not only could such systems bene-

ﬁt from the use of ad-hoc architectures of Ambient In-

telligence, but, in particular, they could be greatly en-

hanced by the use of advanced algorithms from Artiﬁ-

cial Intelligence. Recent studies indicate that the use

of genetic algorithms seems promising (Cook et al.,

2009b; Cook et al., 2009a; ElHillali et al., 2007; Ni,

2007; Zheng et al., 2004; Santos et al., 2010).

In Section 2, the PEGASUS scenario and main

strategies are presented. Section 3 describes the sys-

tem architecture. In Section 4, some drawbacks of

the proposed approach and currently adopted routing

algorithms are detailed and a variant of the core ar-

chitecture is sketched. Such variant is based on virus-

evolutionary genetic algorithms.

2 SMART NAVIGATION

SCENARIO AND STRATEGIES

In Fig.1 the smart navigation scenario considered

is shown: vehicles are equipped with on-board

units (OBUs), which periodically transmit their speed

and position (known through the GPS integrated on

board) to a Control Center. Such data are transferred

through the General Packet Radio Service (GPRS)

network.

Figure 1: Smart navigation scenario.

The ﬂeet equipped with OBUs is addressed as ﬂoating

car data (FCD).

In March 2010, the Italian FCD to which the PEGA-

SUS project refers, reached over 1.000.000 equipped

vehicles (OctoTelematics, 2010); this number is to in-

crease quickly (note that the number of public and pri-

vate vehicles in Italy was 34 million in 2003 (ecoage,

2003), hence the FCD is a not negligible percentage

of the overall private vehicles number).

All such data, once processed, can be exploited for

the real-time dynamic navigation of vehicles or for-

warded to public or privateinstitutions for trafﬁc man-

agement.

The best route has to be calculated and re-

evaluated as soon as possible through a convenient

strategy; the three approaches currently investigated

are the following:

1. Centralized Strategy: evaluation of the best

route at the Control Center;

2. Distributed Strategy: on-board evaluation of the

best route;

3. Hybrid evaluation of the best route.

All the above mentioned strategies are based on

the knowledge of roads conditions on the basis of the

FCD information.

As far as the ﬁrst strategy is concerned, all the data

collected from vehicles are analysed by the Control

Center. A user interested in a given trip asks the Con-

trol Center, which calculates the best route on the ba-

sis of current trafﬁc conditions and transmits it to the

user’s on-board navigator.

In this case, the navigator is a ”dummy” entity, simply

receiving the path.

As for the distributed strategy, the Control Center

periodically transmits the up-to-date road conditions

to all the users.

In this case, thus, the best route calculation is de-

manded to the on-board navigator, which becomes a

AN INTEGRATED ARCHITECTURE FOR INFOMOBILITY SERVICES - Advantages of Genetic Algorithms in

Real-time Route Planning

301

complex and intelligent device.

The hybrid strategy represents a compromise be-

tween the aforementioned solutions. When a user

asks for a route, the Control Center returns updated

trafﬁc conditions so that the on-board navigator can

evaluate the optimum path.

3 THE PROPOSED

ARCHITECTURE

In order to implement the above strategies, it is neces-

sary to design and develop a modular and ﬂexible ar-

chitecture. The proposed one (Fig.2) contains a Con-

trol Center core which manages and processes all the

data collected by the FCD, so as to provide a variety

of infomobility services.

The communication between vehicles (both FCD

and users) and the Control Center is carried out by

two layers: (1) a two-way telecommunication access

network and (2) a user-system interface, the latter

performing operations of format adaptation and con-

tent scalability.

Figure 2: Main Architecture.

The components of the Control Center are detailed

separately in the following:

• Distributed Control Center (DCC): it is the

controller of the whole architecture; when receiv-

ing information (FCD speed and position samples,

best path request, etc.), the DCC forwards it to the

modules in charge of its processing. In the oppo-

site way, the DCC makes the outgoing data (trafﬁc

updates, best path response, etc.) available to the

user-system interface and the telecommunication

access network.

• Dynamic Routing Engine (DRE): it is the rout-

ing engine, and is the module that will be re-

disigned taking genetic algorithms into consider-

ation.

In the current implementation, the DRE deter-

mines the best route applying a Dijkstra-like al-

gorithm to three kinds of information: (1) static,

i.e. the travel time of each road segment in the

absence of trafﬁc; (2) dynamic, based on actual

travel time data measured by the FCD; (3) fore-

cast, i.e. travel time based on trafﬁc forecasts.

All the static, dynamic and forecast information

are stored in speciﬁc databases. The DRE can also

handle Points Of Interest (POIs) along the route,

ﬁnding the optimal path that allows to reach a set

of POIs; this, for instance, can be very useful for

touristic purposes.

• Trafﬁc Control Center (TCC): it is the module

which processes position and speed samples from

vehicles and evaluates the real-time trafﬁc condi-

tions, in particular, the actual travel time needed

to go through a particular road. The TCC also

performs arithmetic, weighted and temporal aver-

age operations in order to estimate the real trafﬁc

conditions of all the segments of the road map.

• Trafﬁc Forecasting Centre (TFC): it is the mod-

ule that analyzes current and historical trafﬁc in-

formation and their trend, forecasting the trafﬁc

evolution over time.

Also this module will be involved in the architec-

ture revision based on genetic algorithms.

All the components described so far make use of

databases storing all the information required for the

processing phases: users, proﬁles, maps, paths, POIs.

The overall architecture in Fig.2 can be ideally di-

vided into four quadrants, as indicated by the different

colours, in order to highlight the four different kinds

of interaction processes developed.

1. The FCD Sampling. Vehicles belonging to the

FCD ﬂeet are equipped with an OBU, so they send

their position and speed to the Control Center

through the telecommunication access network.

2. Best Path Request. On-boardnavigationdevices,

rather than planning routes using their own static

local cartography, could require a real-time short-

est path calculation to the Control Center. In this

kind of interaction, only required for centralized

or hybrid strategies, the on-board navigator sends

a message to the Control Center, setting the cur-

rent position and the required destination, beyond

eventual POIs along the path. Such request is pro-

cessed in Control Center, forwarded by the DCC

to the DRE, in chargeof calculating the best route.

ICEC 2010 - International Conference on Evolutionary Computation

302

3. Best Path Notiﬁcation. This type of interaction

is only required for the on-board strategy. The

Control Center returns a route based on real-time

trafﬁc data, taking into account trafﬁc jams, car

crashes and actual travel times measured by the

FCD. The response given by the DRE is based on

both the data stored in the static Maps database

and all the real-time updates stored in its dynamic

portion, properly integrated by the trafﬁc forecasts

coming from the TFC. The route is returned to

the on-board navigator through a message listing

all the road intersections or milestones needed to

reach the destination.

4. Links Update. This type of interaction is only

meant for the on-board and hybrid strategies. Us-

ing the telecommunication network, it is possible

to send information about updates of the travel

times in an asynchronous way. Each on-board

navigator is thus able to update the road segments

conditions, achieving thus a sort of ”distributed

navigation intelligence”.

By means of this real-time updated information,

transmitted periodically through the telecommu-

nication network, each on-board navigator can

thus apply a routing algorithm which takes into

account the actual trafﬁc conditions, so as to avoid

problematic situations.

4 AN EVOLUTION OF THE

PROPOSED ARCHITECTURE

BASED ON GENETIC

ALGORITHMS

As described in previous sections, the best path in a

given situation can vary when trafﬁc conditions vary

and updates should be notiﬁed to the user in real-time

(Bonnifait et al., 2007; Chen et al., 2007; Jula et al.,

2008; Najjar and Bonnifait, 2007).

If such a feature were enhanced through the use of

appropriate algorithms from Artiﬁcial Intelligence, it

would naturally become an integrant part of advanced

navigators in intelligent environments (Cook et al.,

2009b; Cook et al., 2009a; ElHillali et al., 2007; Ni,

2007; Zheng et al., 2004; Santos et al., 2010).

For instance, the methodologies described so far

could greatly beneﬁt from the use of genetic algo-

rithms instead of traditional search methods. In par-

ticular, on the basis of recent studies on dynamic en-

vironments, a variant of the proposed architecture is

here discussed where virus-evolutionary genetic al-

gorithms replace traditional methods in the Dynamic

Routing Engine and in the Trafﬁc Forecasting Centre.

In order to better explain this viewpoint, some

considerations must be made.

4.1 Exact Routing Algorithms

The solution discussed hereto and all similar ones

arise from classic problems of shortest path ﬁnding

in the static case and try to extend them to real-time

route planning, taking into consideration trafﬁc anal-

ysis, forecasting and path updating.

In particular, the algorithm used so far in the Pegasus

system is a classic variant of Dijkstra. In this kind

of approach, a map is represented by a graph whose

nodes are intersections of roads and whose arcs rep-

resent segment of routes. If no trafﬁc conditions are

considered, arcs are weighted by means of lengths be-

tween nodes. In the dynamic case considered, in order

to represent trafﬁc conditions, length is substituted by

time-varying, actual travel times.

On the basis of data collected about road conditions,

arcs are periodically assigned updated weights and

routes may change consequently.

In context of static environments, shortest path

problems are generally solved in this way, by means

of exact algorithms such as Dijkstra. Many variants

and different approaches were proposed in order to

guarantee better performances, such as A

∗

and many

others, and further variants were also proposed in or-

der to face the real-time case, such as RTA

∗

and PHA

∗

(In-Cheol, 2006; Felner et al., 2004; Korf, 1990). The

variants mentioned above make use of heuristic func-

tions and, step by step, determine a suitable nextmove

until a suboptimal solution is found.

Still, one of the main drawbacks of such ap-

proaches is that, even though adapted to the real-time

case, these algorithms consider only one solution at

a time and do not deal with the entire route until

the end. In consequence, route evaluation is slow;

furthermore, proposing the same alternative to many

clients can give raise to new congestions.

4.2 The Genetic-based Approach

These considerations indicate that the parallel analy-

sis of many solutions could improve the overall per-

formance of the system. In particular, (Kanoh, 2007;

Kumar et al., 2009; Yuecong et al., 2007) suggested

that genetic algorithms could be applied to dynamic

route planning.

With respect to traditional search algorithms, the ge-

netic approach considers many and entire solutions at

a time, so it acquires knowledge and improves the set

of candidate solutions during the search process. The

AN INTEGRATED ARCHITECTURE FOR INFOMOBILITY SERVICES - Advantages of Genetic Algorithms in

Real-time Route Planning

303

efﬁciency of global search improves the limits of tra-

ditional algorithms.

Yet, genetic algorithms applied to dynamic route

planning present a limit: although directed by the ﬁt-

ness function, the stress is on random search rather

than directional, as it should be in order to solve local

trafﬁc problems.

4.3 The Virus-enhanced Variant

An alternative which optimizes the genetic approach

and seems to overcome the above limits was pro-

posed in (Kanoh, 2007) and enhances genetic algo-

rithms through viral mutations. The basic idea is that,

whereas in the static case diversity in the population

is a key factor to reach convergence, in dynamic en-

vironments evolvability is also needed, meant as the

ability of members to change to meet the new require-

ments of the dynamic environment. This feature can

be guaranteed using viral mutations. As a matter of

fact, whereas typical genetic algorithms may not be

able to solve large-scale problems within a practical

amount of time, viruses give a direction to the search,

improving thus search rate, quality of solutions and

speeding the whole process up.

In a general virus-enhanced genetic algorithm

(VEGA), two kinds of population are considered: the

traditional host population of candidate solutions and

a virus population (more properly, a substring set).

First, viruses infect the host population (horizontal

propagation), then viruses are transmitted to offspring

(vertical inheritance).

In more detail, a VEGA includes genetic operators

and virus infection operators, namely reverse tran-

scription and transduction. When reverse transcrip-

tion is applied, a virus transcribes its content on the

string of a host individual; in case of transduction, a

virus transduces a substring from the host individual.

The virus-infection operators deﬁned in this way

are added to the usual selection, crossover and muta-

tion ones.

To some extent, viruses can be regarded as local

changes that can be used to enhance modiﬁcations

in speciﬁc parts of the whole solution space. A part

of a road is considered as a virus and a population

of viruses is generated in addition to a population of

routes. Crossover and infection together determine

the near-optimal combination of viruses. When traf-

ﬁc congestion varies, a better route is determined in

real-time using viruses and other routes in the popu-

lation.

The results in (Kanoh, 2007), which simulates

28.000 cars in Northern Tokio, is that genetic algo-

rithms improve the performance of exact algorithms

Figure 3: Virus-based variant of the Routing Engine and

Forecasting Center.

in both the static and dynamic case and, if further mu-

tations are applied, based on viral infections, the dy-

namic case can be solved even more quickly.

In dynamic route planning environments, fore-

casting models are the premise for developing ur-

ban Intelligent Transportation Systems. In (Yuecong

et al., 2007), a proposal can be found which applies

genetic algorithms to traditional forecasting models.

Our proposal is that virus-enhanced genetic algo-

rithms could be applied to the PEGASUS system in

order to improve its overall performance and, in more

detail, to use viruses to drive the search of better solu-

tions directly where trafﬁc jams and problematic situ-

ations are detected.

Fig. 3 shows how the core of the proposed architec-

ture can be modiﬁed. First, on the basis of data traf-

ﬁc analysis made by the Trafﬁc Control Center, suit-

able viral populations could be deﬁned representing

the most problematic trades (Fig. 3a). This popula-

tion could be used to deﬁne the viral population of a

Virus Evolutionary Genetic Algorithm (Fig. 3b), so

ICEC 2010 - International Conference on Evolutionary Computation

304

as to direct genetic operations in such areas. This al-

gorithm could be directly applied to feed the Trafﬁc

Forecasting Center (Fig. 3c). The trafﬁc ﬂow deter-

mined in this way could be directly used by the Dy-

namic Routing Engine (Fig. 3d).

5 CONCLUSIONS

In this paper an architecture was described for the

constant monitoring of road conditions and the con-

sequent real-time update of routes affected by prob-

lematic trafﬁc conditions.

The system relies on a Dijkstra-likealgorithm and,

since this approach is not suitable to handle dynamic

large-scale problems, a ﬁrst bibliographic research

was carried out in order to compare differentsolutions

to the search of optimal routes in the real-time case.

Some authors indicate that good performances can be

achieved using virus-evolutionary genetic algorithms,

and a variant of the proposed architecture was conse-

quently sketched.

Future work will be devoted to the reﬁnement of the

considered approach and to suitable simulations.

ACKNOWLEDGEMENTS

This work was supported by the PEGASUS project,

ﬁnanced by the Italian Ministry of Economic Devel-

opment.

REFERENCES

Bonnifait, P., Jabbour, M., and Dherbomez, G. (2007).

Real-time implementation of a gis-based localization

system for intelligent vehicles. EURASIP Journal on

Embedded Systems.

Car2Car (2010). Car-to-car communication consortium.

Website. http://www.car-to-car.org/.

Chen, Y., Bell, M., and Bogenberger, K. (2007). Reliable

pretrip multipath planning and dynamic adaptation for

a centralized road navigation system. IEEE Transac-

tions on Intelligent Transportation Systems, 8(1):14–

20.

Cook, D. J., Augusto, J. C., and Jakkula, V. R. (2009a). Am-

bient intelligence: Technologies, applications, and op-

portunities. Pervasive Mob. Comput., 5(4):277–298.

Cook, D. J., Hagras, H., Callaghan, V., and Helal, A.

(2009b). Making our environments intelligent. Per-

vasive Mob. Comput., 5(5):556–557.

ecoage (2003). Independent ecology portal. Website.

http://www.ecoage.net/.

ElHillali, Y., Rivenq, A., Tatkeu, C., Rouvaen, J., and

Ghys, J. (2007). Embedded localization communica-

tion system designed for intelligent guided transports.

EURASIP Journal on Embedded Systems, 19(2):–.

ERTICO (2010). Cvis project. Website.

http://www.cvisproject.org/.

EU (2010). Eu mobility and transport. Website.

http://ec.europa.eu/transport/publications/statistics/.

Felner, A., Stern, R., Kraus, S., Ben-Yair, A., and Ne-

tanyahu, N. (2004). Pha*: Finding the shortest path

with a* in an unknown physical environment. Journal

of Artiﬁcial Intelligence Research (JAIR), 21(1):631–

670.

In-Cheol, K. (2006). Knowledge-Based Intelligent In-

formation and Engineering Systems, chapter Real-

Time Search Algorithms for Exploration and Map-

ping. Springer, Berlin/Heidelberg.

Jula, H., Dessouky, M., and Ioannou, P. A. (2008). Real-

time estimation of travel times along the arcs and ar-

rival times at the nodes of dynamic stochastic net-

works. IEEE Transactions on Intelligent Transporta-

tion Systems, 9(1):97–110.

Kanoh, H. (2007). Dynamic route planning for car navi-

gation systems using virus genetic algorithms. Inter-

national Journal of Knowledge-based and Intelligent

Engineering Systems, 11(1):65–78.

Korf, R. (1990). Real-time heuristic search. Artiﬁcial Intel-

ligence, 42(3):189–211.

Kumar, A. J. S., Arunadevi, J., and Mohan, V. (2009).

Intelligent transport route planning using genetic al-

gorithms in path computation algorithms. European

Journal of Scientiﬁc Research, 25(3):463–468.

Najjar, M. and Bonnifait, P. (2007). Road selection us-

ing multicriteria fusion for the road-matching prob-

lem. IEEE Transactions on Intelligent Transportation

Systems, 8(2):279–291.

Ni, D. (2007). Determining trafﬁc-ﬂow characteristics by

deﬁnition for applications in its. IEEE Transactions

on Intelligent Transportation Systems, 8(2):181–187.

OctoTelematics (2010). Octotelematics s.r.l. Website.

http://www.octotelematics.it/.

PEGASUS (2010). Pegasus project. Website.

http://pegasus.octotelematics.com.

Santos, A. C., Cardoso, J. M., Ferreira, D. R., Diniz, P. C.,

and Chainho, P. (2010). Providing user context for

mobile and social networking applications. Pervasive

Mob. Comput.

UN (2008). World urbanization prospects: The

2007 revision population database. Website.

http://esa.un.org/unup/.

Yuecong, S., Wei, H., and Guotang, B. (2007). Combined

prediction research of city trafﬁc ﬂow based on ge-

netic algorithms. In Proc. of the Eighth International

Conference on Electronic Measurement and Instru-

ments ICEMI 2007.

Zheng, N., Tang, S., Cheng, H., Lei, Q. G., and Wang, E.

(2004). Toward intelligent driver-assistance and safety

warning systems. IEEE Intelligent Systems, 19(2):–.

AN INTEGRATED ARCHITECTURE FOR INFOMOBILITY SERVICES - Advantages of Genetic Algorithms in

Real-time Route Planning

305