An Autonomic System for Intelligent Truck Parking

Jose F. Garc

ıa

, Vicente R. Tom

, Luis A. Garc

ıa

and Juan J. Mart

ınez

Institute of robotics and information and communication technologies (IRTIC), University of Valencia, Paterna, Spain

Engineering and Computer Science Department, University Jaume I, Castell

on de la Plana, Spain

Keywords:

Autonomic Systems, Multi-agent Systems, Intelligent Truck Parking, Negotiation Systems.

Abstract:

Trafﬁc and road transport conditions are strongly inﬂuenced by the decisions of drivers. The more informa-

tion they receive, the better decisions may be taken about their behavior. This is especially important in the

transport of goods, since drivers are subject to trafﬁc laws about maximum driving time and minimum time

rest. In this paper a multi-agent system for negotiated management of parking spaces in road rest areas is

presented. This system dynamically adapts itself to the preferences and needs of the drivers of goods about

parking requests. The system is shown to be robust to incidents regarding the closure of road rest areas and

an increased volume of freight trafﬁc. The results also show that the number of illegal parkings is reduced

resulting in greater road safety.

1 INTRODUCTION

The domain of road trafﬁc and transport involves an

inherently distributed environment. Several trafﬁc or-

ganizations must act in a coordinated way to improve

the trafﬁc safety and status, for example by identify-

ing trafﬁc ﬂow problems and by proposing actions to

alleviate them. This coordination is fundamental in

Long Distance Corridors (LCD). A LDC is a road net-

work corridor that includes a common platform with a

co-ordinated setup of different systems and services,

contributing to improve trafﬁc management and in-

formation for travelers and freight transport on long

distances (Aumund, 2004). The main use of these

corridors are the transport of goods. Therefore, route

planning plays an important role in the goods road

transport domain. However, route planing does not

only include the itinerary but also the rest road ar-

eas to park and rest because journey time exceeds the

maximum driving time required by law. Moreover, if

drivers park in not allowed places, they can produce

problems in trafﬁc safety and be ﬁned. The impor-

tance of the parking spaces in Europe is reﬂected in

several institutional or research actions: ITS direc-

tive (Commission, 2010) , the EU ITS Action Plan

(Commission, 2008) and the deployment guideline

for Intelligent truck parking developed in the Easy-

Way project (EasyWay, 2012).

An Intelligent Truck Parking system (ITPs) is de-

ﬁned as an ITS service that support drivers in two

ways: a) information and guidance (on truck parking

areas) and b) reservation (of truck parking spaces).

ITPs could be classiﬁed attending the level of Ser-

vice (LoS). LoS deﬁnes the speciﬁc characteristics

that the ITP offers to the drivers. In (EasyWay, 2012),

5 different levels are deﬁned, from A (lower level) to

E (high level). Currently, several organizations and

projects have developed ITPs projects regarding level

A to C. These systems are based on static informa-

tion, based on books and journey routes periodically

edited (ParkMyRig, 2014) (Project SETPOS, 2014).

Regarding level of service D, systems with dynamic

information, there are also different systems (Sndor

and Csiszrr, 2013) (Vennekens, 2014). These book-

ing systems are based on pre-trip systems, i.e. the

driver has to book the place to rest before to begin the

journey.

In (ParkYA, 2014), a new application for mobile

devices has been developed to make easier to ﬁnd car

parking spaces in real-time. Despite it is a good ap-

proach, its main problem is that only drivers use it, the

parking areas are not involved in the system. Thus,

the parking availability is deﬁned by drivers, that in-

form, in a qualitative and subjective way, the occu-

pancy of areas.

So, current systems present several difﬁculties be-

cause drivers do not know the trafﬁc behavior in ad-

vance, so they might even not reach the parking area

booked, or when they reach it the area could not have

free places. Therefore, an ITP system where drivers

810

Garcia J., Tomás López V., García Fernández L. and Martínez Durá J..

An Autonomic System for Intelligent Truck Parking.

DOI: 10.5220/0005149008100816

In Proceedings of the 11th International Conference on Informatics in Control, Automation and Robotics (IVC&ITS-2014), pages 810-816

ISBN: 978-989-758-040-6

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

snowzone

carry

chains

snowzone

carry

chains

Truckagent

TISagent

Areaagent

TISagent

snowzone

carry

chains

snowzone

carry

chains

TruckagentTruckagent

Truckagent

Reservationandnegotiationprotocols

Informationprotocols

RealWorld

MASSystem

Figure 1: An example of the system architecture. At the top, a real world example is presented, including trucks, variable

message signs and a rest area. On the bottom, the agents that compose the system are showed. Each truck has its own agent,

there is also an agent for VMS and one agent for each area.

could modify their reservation on-trip, depending on

the areas availability, are required. By this way, there

is a demand for ITS services where the real-time com-

munications between vehicles and infrastructure. In

the last years, the advances of the communication

technology have made possible the deployment of co-

operative systems. A cooperative systems is a sys-

tem where the different elements involved in the traf-

ﬁc domain can communicate with each other. These

cooperative systems are usually addressed regarding

the type of elements involved in the communications:

V2V vehicle to vehicle, V2I vehicle to infrastruc-

ture, I2I infrastructure to infrastructure, etc. (Alexan-

der Paier, 2013). But, not only cooperation is impor-

tant, also the way this cooperation enables to run traf-

ﬁc objective functions, even in scenarios in which not

central control is possible as it happens in the trafﬁc

domain. In (COST TU1102, 2015) new ways of de-

signing Road Transportation Support (RTS) systems

based on the ideas of autonomic systems are being

analyzed. The concept of autonomic computing was

deﬁned at the end of the XX century (Kephart and

Chess, 2003) and it deﬁnes an autonomic system as

a system able to self-manage without the need of an

external or central manager that identiﬁes the action

to be deployed in each situation or scenario.

In this paper, a new ITP system with self manage-

ment properties is presented. It is deﬁned as a mul-

tiagent architecture following a V2I approach. The

system imposes the coordination between the trafﬁc

elements involved in the solution for an efﬁcient oc-

cupation of the rest areas. It uses a new negotiation

protocol suitable to adapt the dynamically required

places by the trucks to the available places. Besides,

the data exchanged for this negotiation uses the stan-

dard trafﬁc data, DATEX.

The paper is structured as follows. Next section

presents the proposed system for ITP. It describes

the architecture, including the knowledge and infor-

mation models and the communications and negotia-

tion protocols. Then, in Section 3, the real network

is modeled and the different experiments to test and

evaluate the system are described. Finally, the con-

clusions are exposed.

2 A DISTRIBUTED SYSTEM FOR

ITP

The objective of the system is to support truck drivers

to ﬁnd a place in the most preferable parking area ac-

cording to their preferences. This system will also im-

prove the parking management, because it can man-

age, in real time, the real occupancy of its parking

places, avoiding bad parking nearby the area.

The system is based on the multiagent (MAS)

paradigm. The deﬁnition of this proposed MAS con-

tains: the architecture, the description of agents, the

information and knowledge model, both public and

private for each agent and the negotiation protocol to

manage the reservation of free parking spaces.

AnAutonomicSystemforIntelligentTruckParking

811

2.1 System Architecture

The MAS is composed by several kinds of agents:

area agent, truck agent, TIS agent, DF agent and inter-

face agent. Figure 1 shows the software architecture

of this MAS prototype.

2.1.1 Area Agent

Area agents are in charge of managing parking areas.

Areas are located along the road network and have a

limited number of parking spaces. The agent is re-

sponsible of several tasks:

• to attend the reservation process when places are

available.

• to execute the negotiation protocol when there are

more reservations than available places.

• to exchange information about space availability.

2.1.2 Truck Agent

Vehicles circulate along the road network. Truck

agents are responsible of individual truck vehicles.

Using the current truck location, the mean speed, and

the driving times, it forecasts the possible next areas

where the driver could park to rest. These areas are

selected following the preferred order of the driver.

2.1.3 TIS Agent

The Trafﬁc Information Service agent receives infor-

mation from the different parking areas. It manages

a set of Variable Message Signs and it diffuses in-

formation about the places availability in the parking

spaces. This agent is an extension of the VMS agent

described in (Tomas et al., 2012)

2.1.4 DF Agent

The DF agent provides a yellow pages service. Each

agent can register, deregister and search for other

agents or services available in the MAS platform or in

other platforms. The directory facilitator used is the

DF JADE agent (TILAB, 2014) speciﬁed by FIPA.

2.1.5 Interface Agent

The goal of this agent is to show how the system sta-

tus evolves dynamically. This agent is based on a

web application and it displays the road network and

the reservation status via Google Maps. The map has

the normal controls of a Google Maps plus a speciﬁc

menu to show all the areas and trucks in the road, in-

cluding their public information.

2.2 Information and Knowledge

The information and knowledge model is structured

in two levels: a) the road network model describing

the elements composing the road, including the park-

ing areas and the truck based on DATEX II standard

(DATEXII, 2014) and b) the reservation and negotia-

tion process information.

2.2.1 Road Network Model

This domain is based on (Martinez et al., 2013). It

has been extended and is composed by the following

elements:

• Segments. They represent two way road sections

and their characteristics (road name, number of

lanes, length, free-ﬂow speed, capacity, trafﬁc re-

strictions, ...).

• Links: They represent the road network elements

where two, or more, adjacent segments are con-

nected. The link objects are subdivided in: bifur-

cations, unions, weavings and merges.

• Areas. They represent the rest areas where vehi-

cles can park. The main characteristics are: seg-

ment, kilometer point (kp), facilities, truck capac-

ity, current occupancy and list of reservations.

• Trucks. It models the trucks circulating by the

road network. The main characteristics are: cur-

rent location (segment and kp), origin, destina-

tion, maximum speed and current driving time.

2.2.2 Reservation and Negotiation Model

• Reservation. It represents an area choice from a

speciﬁc truck. It includes the truck, the area, the

expected arrival time according trafﬁc conditions

and the expected permanence time.

• Place occupied. It represents a parking place oc-

cupied by a truck. It includes the truck, the area,

the time the truck parked and the permanence

time.

• Preference. It represents the truck’s driver prefer-

ence to a speciﬁc area.

• List of preferences. It is a ordered list identifying

the preferences of a speciﬁc driver.

• Assignation. An assignation is a pair of elements

< truck

, area

> indicating that the truck a has

been assigned to an area i.

• Feasible Solution. A feasible solution is a tuple

of Assignations. It includes a possible assignation

for the set of trucks and areas.

ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics

812

• Vote. It represents the score for a speciﬁc Feasible

Solution.

• Final Solution. It represents the negotiation solu-

tion. It is the feasible solution that obtained the

higher score.

2.3 Communication Protocols

The communication protocols deﬁne the way agents

are related to each other in the platform to exchange

information or to register other agent services. The

communication protocol proposed follows the inter-

action protocols deﬁned by FIPA (Foundation for In-

telligent Physical Agents, 2000). Next, the main sys-

tem protocols are described. The basic protocols like

the registration to DF and similar are not deﬁned.

• Reserve. Request protocol. The truck agent sends

to an area a message to obtain a reservation. If

the protocols ends successfully the area answers

with a reservation. If not, a negotiation protocol

between the area and the trucks begins.

• Ask list preferences. Query protocol. The area

agent ask to the different trucks involved in the

negotiation their preferences.

• Send preferences. Inform protocol. List of pref-

erences sent by trucks to the area agent managing

the negotiation.

• Conﬁrm reservation. Inform protocol. Conﬁrma-

tion sent to trucks involved in the winning solu-

tion.

• Disconﬁrm reservation. Inform protocol. The

area agent informs to the truck about the cancella-

tion of the requested reservation.

• Park Status register. Request protocol. The TIS

agent sends a message to be registered in the area.

The message contains the occupancy rate. When

the area reaches this percentage, it informs the TIS

agent registered.

• Park Status inform. Inform protocol. The area

sends a message to the registered TIS agents when

the occupancy rates have been reached.

2.4 Negotiation Protocol

The negotiation protocol is launched when an area has

more reservations than available parking places. The

area agent communicates with all involved trucks to

begin the negotiation. The negotiation will be per-

formed using a variation of Borda voting based proto-

col (Capdevila et al., 2013).

The negotiation begins and the area asks to all

trucks their list of preferences. Each truck involved

in the negotiation will assign a value to each of its

preferences. In this approach, the truck preferences

are generated taking into account only the time that

the truck is on the network without stopping, and the

distance to the areas. The preferences ordering is au-

tomatically calculated by means of an heuristic that is

very close to a normal truck driver behavior: drivers

try to maximize driving time without stop stabilized

by law, i.e., the most remote area reachable for the

truck will be the most preferred and the closest area

will be the least preferred area.

Once the area has received all trucks preferences,

it generates the set of feasible solutions which is com-

posed by the set of possible assignations of trucks in

the areas involved in the solution. These feasible so-

lutions are sent to the trucks. Each truck calculates

its votes and then, they are sent to the area. The area

calculates the result and conﬁrms the reservation to

the trucks that have the reservation and disconﬁrms

the reservation to the truck that is not in the winning

solution. This truck will ask for a reservation in its

following preferred area.

2.5 System Implementation

The system has been implemented in JAVA using a

JADE (TILAB, 2014) platform. JADE is a software

framework to develop agent applications in compli-

ance with FIPA speciﬁcations for interoperable intel-

ligent multiagent systems. Figure 2 shows a snapshot

of the interface agent.

Figure 2: Interface agent snapshot. A section of the sys-

tem is presented. Two parking areas, one without parking

places are showed. The information provided via VMS is

also presented.

AnAutonomicSystemforIntelligentTruckParking

813

3 THE MEDITERRANEAN

CORRIDOR: E-15 MOTORWAY

To analyze and evaluate the proposed MAS, a real

road network has been modeled. This road network

is part of the Spanish Transeuropean Road Network

(TERN) and belongs to one of the main european cor-

ridors, the E-15 Long distance corridor. It is used to

transport goods not only from Spain but also from the

North of Africa to Europe.

The modeled part covers the 400 km. of the Va-

lencian Community. The elements modeled includes:

• 1 main road, E-15 (AP-7, A-7 & A-70 national

motorways)

• 27 Segments.

• 25 Links.

• 24 VMS.

• 11 storage areas.

Three different scenarios have been developed to

analyze the system: a) the system behavior in a real

situation, b) the system robustness and c) the system

scalability.

A set of 100 simulations have beed developed for

each scenario. The areas and their availability are the

same for all simulations. Truck positions have been

created randomly. In all simulations, trucks drivers

have been driving for some time, so, most of them

will have to park in a rest area in the Valencian Com-

munity. The information of trucks and areas for the

100 simulations have been maintained for the three

scenarios.

3.1 A Real Situation Test

In this scenario, the real trafﬁc on heavy good vehi-

cles have been simulated. The information is based

on the Annual Average Daily Trafﬁc (AADT) for the

year 2012. (Fomento, 2012). The AADT varies little

except in the ring-roads areas of the main cities. In

the ring-roads, the intensity is too high because of the

trafﬁc of goods around cities. This intensity has not

been taken into account because drivers do not drive

long distances. In the E-15, the truck AADT is close

to 1100 veh. The truck intensity in this scenario has

been calculated using the trafﬁc ﬂow distribution in

the E-15 during the day and the time used by drivers

to cross the scenario. So, the intensity used in this

scenario is deﬁned to 400 trucks. The capacity of all

areas has been harmonized to 40 places. This has been

done to make easier the results analysis. The results

are presented in ﬁgure 3.

Results show how without negotiation there are a

lot of trucks parking in areas without places and they

have to park on ramps, hard shoulders... This situa-

tion is specially difﬁcult in La Marina. These situ-

ations happen because drivers want to reach the far-

thest area. However, applying the negotiation proto-

col, trucks are distributed in the upstreams areas (Sa-

gunto, Picassent and La Safor). In these areas there

are also truck parking without places, but this num-

ber is decreased a lot. This situation happens due to

the random location of trucks and their driving times.

There are no optimal solution where all vehicles have

a parking area.

3.2 Testing the Robustness

To assess the system robustness, the baseline scenario

has been compared with a new scenario where one of

the areas (concretely La Safor) had a problem and it

can not store trucks. This situation can be caused by

different problems: an incident in the area that does

not allow trucks to access it, or communication prob-

lems and the area could not negotiate its reservations.

Figure 4 presents the results of the same simula-

tions but including the simulation without La Safor

area. The results shows how applying the negotiation

trucks are distributed in the area upstreams. However,

without negotiation trucks park in the previous area,

Picassent. This situation is strongly presented in sim-

ulation 14. It is worth to note from these results that

the behavior of the areas far away to the area cancelled

out is quite similar with respect to their behavior when

all areas are available.

So, the system is robust since in spite of the fail-

ure in one of the areas, the system continues working

and trucks start to negotiate with the following areas

according to their preferences.

3.3 Testing the Scalability

To assess the system scalability, the number of trucks

and the number of parking spaces in the areas have

been increased up to 1000 vehicles and 100 parking

places in each area.

The run-time and the number of messages ex-

changed increase linearly. So, there are no problems

to execute the platform and agents in real time. How-

ever, before to conclude the system scalability results

have to be analyzed.

Figure 5 presents the result of increasing the num-

ber of trucks. As in previous scenarios, when the ne-

gotiation protocol is used, trucks try to distribute the

parking in their preferred areas, avoiding, if possible,

ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics

814

Figure 3: Results of simulation 6 and 14 are presented. The results of parking assignments with and without negotiation are

presented.

Figure 4: Results of simulation 6 and 14 comparing previous scenario and a new one dropping La Safor.

to exceed the areas capacity. This situation does not

happen when the negotiation is not used.

Figure 5: Simulation results increasing the number of trucks

to 1.000.

4 CONCLUSIONS

In this paper, a multiagent system implementing a

new ITP system is presented. The system uses a ne-

gotiation protocol to coordinate the parking of trucks

in rest areas. The ITP protocol tries to guarantee that

all vehicles have a parking space when they reach the

rest area. In this way drivers comply laws about driv-

ing time and rest periods.

This new system improves the current ITP sys-

tems because it takes into account not only the current

parking places but also the preferences of drivers. So,

drivers can know in advance if they have a parking

place reserved.

The negotiation protocol is adapted to the dy-

namic behavior of the trafﬁc. When an area has no

parking places, trucks trying to park in this area start a

negotiation using votes. Each truck votes his best op-

tion from a possible set of solutions. The voting pro-

tocol is a modiﬁcation of the Borda protocol adapted

to this speciﬁc problem.

Negotiation processes and speciﬁcally voting sys-

tems always have the possibility of manipulation ei-

ther by coalitions, previous knowledge... In this pa-

per, it is assumed that drivers involved in the negotia-

tion are benevolent. So, if some truck is assigned to an

area that is not his ﬁrst preference, it parks there. The

system has not been designed from the scratch. Some

agents and the trafﬁc domain ontology have been ex-

tended from previous work. Furthermore, it uses the

FIPA and DATEX II communication standards, so it

is compatible with other existing trafﬁc systems.

Results are very positives and promising. Three

different scenarios were developed to analyze the sys-

tem. In the ﬁrst scenario, a real situation has been

tested. In all 100 simulations, the results show that

using the system and the negotiation the number of il-

legal parking places is reduced.

In the second scenario, the robustness of the sys-

tem has been tested. To do it, an area was dropped

from the scenario and the results, both using and not

using the negotiation, were compared with the results

of the ﬁrst scenario. In this situation the illegal park-

ing increases due to the absence of a parking area (40

places less than before). However, using the negoti-

ation the solution improves the results, indicating the

system has the capacity and the ability to deal with

AnAutonomicSystemforIntelligentTruckParking

815

failures in any of its elements.

In the third scenario, the scalability has been

tested. 1000 trucks were in the road. The system

running behavior is not affected in a sensible way.

The run-time and the increase of messages exchanged

does not affect the system. Furthermore, the results

using the negotiation were better again than the re-

sults without using it.

The self management autonomic property of the

system has been demonstrated. This property has

been reached in two levels: a) at individual level,

areas manage their capacity and prevent the excess

of parking and trucks are guaranteed to access to

their preferred area; and b) at network level, although

trucks deal on their preferred areas, the overall sys-

tem performance is greatly improved, since trucks

have information in real-time and illegal parking is

reduced.

Despite the positive results, there are still work to

do. Currently, an improvement of the negotiation pro-

tocol is being developed. This improvement consists

in using the duration of the rest and the speed of trucks

perform a small negotiation between pairs of trucks in

the area, instead of begining a new negotiation among

all trucks with a reservation.

ACKNOWLEDGEMENTS

This work has been supported by a Universitat Jaume

I-Fundaci

o Caixa-Castell

o research project number

P11B2011-46.

REFERENCES

Alexander Paier, Reﬁ-Tugrul Guner, W. B. (2013). V2X Co-

operative systems On the way to next generations ITS.

Prace Naukowe Politechniki Warzawskiej, Vienna.

Aumund, H.-J. (2004). VIKING Long Distance Corri-

dor Feasibility Study. VIKING TEMPO Programme

MIP03, Hannover.

Capdevila, M., Tom

as, V. R., Garc

ıa, L. A., and Prades, M.

(2013). Dynamic management of parking spaces with

automatic negotiationt. IEEE International Confer-

ence on Systems, Man, and Cybernetics (SMC).

Commission, E. (2008). Action plan for the deploy-

ment of Intelligent Transport Systems in Europe.

COM/2008/0886. EU, Brussels.

Commission, E. (2010). Directive 2010/40/EU of the euro-

pean parliament and of the council of 7 July 2010 on

the framework for the deployment of Intelligent Trans-

port Systems in the ﬁeld of road transport and for in-

terfaces with other modes of transport. EU, Brussels.

COST TU1102 (2013-2015). TUD COST Action TU1102,

Towards Autonomic Road Transport Support Systems.

www.cost-arts.org. http://www.cost-arts.org. Vienna.

DATEXII (2014). www.datex2.eu.

EasyWay, P. (2012). DG-FLS-DG01.-Intelligent Truck

Parking and Secure Truck Parking. Deployment guide-

line. EW-DG, Brussels.

Fomento, M. (2012). Heavy good vehicles and hazardous

goods. Trafﬁc Map. http://jade.cselt.it, Madrid.

Foundation for Intelligent Physical Agents, F. (2000). Fipa

interaction protocol library speciﬁcation.

Kephart, J. and Chess, D. (2003). The vision of autonomic

computing. computer. Computer, 36(1).

Martinez, J., Tomas, V., Martinez, J., and F., S. (2013).

Tmeteosafety: A multiagent system to support the au-

tomated activation of trafﬁc management plans for ad-

verse weather situations. TRB Annual Meeting.

ParkMyRig (2014). The Ultimate Truck Parking Guide -

3nd Edition. http://www.parkmyrig.com/Truck Park-

ing Guide.html. Visited June 2014.

ParkYA, T. (2014). ParYA app. http://parkya.com. Dublin.

Project SETPOS (2014). The european truck information

Portal. http://www.truckinform.eu.

Sndor, Z. P. and Csiszrr, C. (2013). Development Stages

of Intelligent Parking Information Systems for Trucks.

Acta Polytechnica Hungarica.

TILAB (2014). Java Agent Development framework

(JADE). http://jade.cselt.it.

Tomas, V., Marti, I., Saez, A., and Martinez, J. (2012).

Agent-based test-bed for road information systems.

Intelligent Transport Systems, IET, 6(4):404–412.

Vennekens, B. (2014). Parkr system. DATEX User Forum,

Prague.

ICINCO2014-11thInternationalConferenceonInformaticsinControl,AutomationandRobotics

816