© Springer International Publishing Switzerland 2015
J. Bajo et al. (eds.), Trends in Prac. Appl. of Agents, Multi-Agent Sys. and Sustainability,
65
Advances in Intelligent Systems and Computing 372, DOI: 10.1007/978-3-319-19629-9_8
Integration of a Purchase Porter System into a VO
Developed by PANGEA
Katsumi Harashima1, Hiroki Nakazaki1, Pablo Chamoso2,
Javier Bajo3, and Juan M. Corchado2
1 Osaka Institute of Technology, Osaka, Japan
2 Computer and Automation Department, University of Salamanca, Spain
3 Department of Artificial Intelligence, Technical University of Madrid, Spain
{harashima,e1611071}@oit.ac.jp,
{chamoso,corchado}@usal.es, jbajo@fi.upm.es
Abstract. Practical applications of multi-agent systems have acquired a grow-
ing relevance during last years. This kind of system is based on distributed arti-
ficial intelligence and they pay special attention to organizational concepts, This
allows a more realistic design of artificial societies. In this paper, we are pre-
senting a multi-agent system based on virtual organizations, which has been
specially designed to model a Purchase Porter System aimed at helping clients
of shops to transport their shopping. The system establish agreements to trans-
port shopping to the more appropriated exists for the client. The system has
been tested in a simulated environment and the initial results are promising.
Keywords: purchase porter system, multi-agent systems, virtual organizations.
1 Introduction
Even though e-commerce is the most popular commercial method over the last few
years, most consumers still go shopping at commercial centres.
A large part of the success of these shopping centres is based on the wellbeing of-
fered to clients. According to [1], six are the main factors that affect comfort: func-
tionality, convenience, safety, leisure, atmospherics and self-identification.
However big malls have disadvantages. For example, clients usually report they
find difficult taking bags and shopping items from one shop to another. Obtaining a
solution for this problem requires designing new mechanisms for the acquisition and
management of huge amounts of information about the environment.
Multi agent systems (MAS), used as a way to create artificial societies, have
achieved great importance during the last years [2][3].
MAS are one kind of distributed artificial intelligence in which autonomous and
individual agents collaborate and cooperate within an artificial society to achieve
defined objectives. These MAS incorporate advanced reasoning models as well as
learning and adaptation capacities.
One of their many characteristics is the ability to represent interaction models,
close to the ones found in human societies [4]. Nonetheless, they do not consider
organizational aspects such as rules or regulations.
66 K. Harashima et al.
Therefore, it is necessary to utilize virtual agent organizations to implement a real-
istic artificial society. These organizations allow to represent regulatory and manage-
ment aspects, which is a much better implementation, being a more efficient and prac-
tical solution.
In the framework of this paper, it is proposed a help system to aid customers in
shopping malls, designed as a MAS based on virtual organizations (VO), in which
every virtual organization represents each different area of the commercial centre. At
least, one porter agent will exist in each organization and it will be responsible for
receiving all customer purchases, as well as delivering them in a convenient place at a
scheduled time.
To accomplish the implementation of this system, PANGEA [5], an agent plat-
form, has been chosen as it was developed by the participants of this work. This is
important, as it was mentioned before, for resembling the system architecture to the
real structure of shopping malls. Although, virtual organizations are self-adapting
systems that suit specific situations automatically, without needing human interaction.
Through this article, the problem described in section 2 will be explained with
more detail. In chapter 3, the suggested solution, based on virtual organizations, is
introduced and then, its results and conclusions in section 4.
2 Review of the State of the Art
We refer to MAS when two or more agents are able to collaborate for solving a prob-
lem [6]. Concretely, one of them is required to be autonomous, and at least, there
must exist a relation between two agents in which one of those satisfies the objectives
of the other.
The main characteristic is that every agent focus in its own individual task, alt-
hough the data they use is decentralized across the system. So, each agent must estab-
lish their goals and plans to meet their objectives, always taking into consideration the
environment where it is located. However, to accomplish those tasks, agents have to
work coordinately, so this is why the global system needs to implement communica-
tion and control mechanisms.
In order to implement global system control, modelling its structure as a set of VOs
can be very helpful. VO are sets of individuals and institutions that need to coordinate
their resources and services within some institutional bounds [7]. Therefore, agent-
based virtual organizations hold a social structure and a set of rules which regulate the
interaction among the different agents. In addition, they have to own a coordination
mechanism to determine the way agents fulfill their tasks.
There are several previous works which focus their case studies in the application
of models with agent-based virtual organizations to shopping malls, i.e. the one ex-
plained in reference [8]. In that case, a MAS was responsible of managing and plan-
ning routes to clients making use of THOMAS architecture [9].
Integration of a Purchase Porter System into a VO Developed by PANGEA 67
3 A VO-Based Purchase Porter Multi-agent System
In the last few years, many large-scale commercial institutions have found important
business opportunities by creating shopping malls with a great number of brand stores
and services such as cinemas, restaurants or skating rinks. This is one of the reasons
why malls are successful.
When people go shopping at commercial centres, they buy products in different
stores, carrying more and more items from one shop to another. As purchases in-
crease, customers feel tired and uncomfortable.
For the purpose of solving the problem mentioned above, we have proposed a pur-
chase porter system with multi-agent technology. In this porter system, individual
porter agents have their own workspace. They receive purchases from each store and
then, they carry them to the exit to be on time for the desire collection time.
When a porter agent of a certain area receives a transportation request from a shop,
products are delivered from the shop location to the agent position. If clients desire to
pick the items at the exit of this area, the porter agent transport them to the specified
place on time.
Otherwise, the agent communicates with the representatives of adjacent areas. In-
ternally, the best possible route is calculated considering the destination, time remain-
ing and the priority of other requests.
This way, products are optimally carried across several areas by delegating the
transportation task among several porter agents, each one responsible of its own
workspace.
Fig. 1. Transfer of purchase
Figure 1 shows the transfer of purchases among agents. In Fig 1(a), the initial sit-
uation is represented, having the agent A, a purchase needed to be delivered at the
destination point. As this location is at the exit of agent B workspace, A gives the
purchase to agent B, Fig 1(b).
With this system, clients do not need to take items with them every time they enter a
new shop. Also, purchases are collected whenever the client wants and in the exit they
choose. Usually, customers would like to pick all the items when they return home.
68 K. Harashima et al.
3.1 Features of the Purposed Porter System Are as Follows:
3.1.1 Environment
The environment has several stores and exits.
• Store
In each store, transportation of purchases occurs constantly.
• Destination
Every destination is located on the outer circumference of the environment and
they become the final destination of the purchase.
3.1.2 Purchase
Every purchase has information about a collection point (exit) and time, both set by
the customer. Our system adds a priority calculated according to the remaining time
and distance to the destination.
Table 1 shows priority for the destination time. Our system attaches a priority so
the system is working full time. The shorter the remaining time, the higher the priority
is.
Table 2 shows priority for the distance to the destination. It includes a priority sys-
tem so that the distance becomes as long as possible. When a distance is shorter, the
associated priority to the distance is larger, due to efficiency.
Porter agents are able to transport purchases by evaluating these priorities without
being late for a time limit.
3.1.3 Agent
An individual porter agent has a certain workspace and it can only move in its own
area. They receive the purchases of a shopper from each store. Then they carry those
purchases to the destination to be ready for collection on time.
Table 1. Priority for remaining time
Table 2. Priority for distance
Integration of a Pur
Each agent behaves as desc
1. First, it receives infor
shop in its workspace.
2. It compares the new pu
3. If the new priority is
mation and receive the
4. If the destination of th
workspace, then it take
5. If a new purchase in its
6. If the adjacent agents
those agents only if the
7. If an agent does not ha
and then starts again fro
3.2 Integration in a Mu
To implement this system
PANGEA as it is based in
the real world, is going to p
ferent structures shopping m
organizations.
Fig. 2. Image showing the
chase Porter System into a VO Developed by PANGEA
ribed below:
mation about a new purchase (transport request) from
rchase priority with the others queued.
higher, it has to move to the shop which sent the inf
items. Otherwise, it does not do anything.
e purchase with the highest priority is within the agen
s it there. Otherwise, jump to step 6.
workspace is received, jump to step 1.
are located near the destination, it gives the purchases
requests in their queues have lower priorities.
ve any purchase, it moves to the center of its worksp
m step 1. If not, it jumps to step 5.
lti-agent Platform (PANGEA)
, it has been decided to use multiagent architect
organizational concepts. This, regarding its similarity w
ermit modelling and implementing the large range of
alls have. Although, as a rule, they usually share sim
'Monitoring tool' with the basic agents provided by PANGEA
69
a
or-
t’s
to
ace
ure
ith
dif-
ilar
70 K. Harashima et al.
This characteristic allows to improve the control of each node in the system in an
individual or collective way.
Furthermore, this is fundamental when controlling and monitoring the state of the
client purchases in every moment. This process is accomplished by the ‘Monitoring
Tool’ which allows to keep and visualize records of everything that has happened in
the system. It also enables to define a set of rules which are fundamental during the
process of implementation.
Several specifically designed and structured agents have been added to control
agents, which the architecture presents for using basic functions. These new agents
develop the case study introduced in this article.
Fig. 3. Scheme showing the subsystem developed
In figure 3, it is shown a scheme representing the subsystem structure for the case
study proposed. In it, there are three virtual organizations as an example, but there
will be as many as areas defined in the shopping mall plus one. New areas are defined
by creating new organizations. This task is performed by the organization agent
which is provided by PANGEA.
In each organization associated to an area, there must exist an agent for every shop
existing in that area. This agent will notify the shopping agent each time a new pur-
chase is completed in the shop it represents.
Also, every shopping agent is located in a virtual organization together with the
appointment agent, which is in charge of managing all system priorities as it has been
explained before.
In order to check whether which areas (represented as organizations in PANGEA)
are adjacent among them, several XML rules are defined. They indicate which organ-
izations can communicate among them (adjacent organizations) and those which
Integration of a Purchase Porter System into a VO Developed by PANGEA 71
cannot. This process is carried out by the PANGEA rule agent, using its monitoring
tool following the scheme below:
10
1
1
…
This would mean, as expressed in the rule with identifier equal to 0, that no organi-
zation could communicate with others, unless there would be rules with higher priori-
ty (lower level) indicating the opposite. This is the case for rules with id 1 and 2
which express that organizations (areas) A-B and A-C are adjacent.
4 Results and Conclusions
This article introduces a new porter system with multi-agent systems and virtual or-
ganizations. The main focus of this article is to provide a customer-driven solution in
large scale commercial institutions. However, satisfying all clients may be a compli-
cated task, due to its high number. Despite this, the system is able to transport almost
all purchases to the exit on the specified time. This is possible thanks to the proposed
priority queue.
However, in our system, the stores request delivery of purchases at a constant
interval. But this is unrealistic. Therefore, our future work is to implement several
improvements to address a real flow of purchases over time.
Regarding the use of virtual organizations, the architecture is organized in a way
that resembles a real shopping mall. This way, areas and agents are distributed easily.
Likewise, PANGEA features fit perfectly to create organization-based structures,
permitting control and monitoring all moves occurred. This architecture presents
an adaptive and scalable design capable of adjusting itself to new conditions, being
applicable to any commercial setting.
72 K. Harashima et al.
References
[1] El Hedhli, K., Chebat, J.C., Sirgy, M.J.: Shopping well-being at the mall: Construct, ante-
cedents, and consequences. Journal of Business Research 66(7), 856–863 (2013)
[2] Lane, J.E.: Method, Theory, and Multi-Agent Artificial Intelligence: Creating computer
models of complex social interaction. Journal for the Cognitive Science of Religion 1(2),
161–180 (2014)
[3] Fan, Z., Duan, W., Chen, B., Ge, Y., Qiu, X.: Study on the method of multi-agent genera-
tion algorithm within special artificial society scene. In: 2012 UKACC International Con-
ference on Control (CONTROL), pp. 1076–1081. IEEE (September 2012)
[4] Schreinemachers, P., Berger, T.: An agent-based simulation model of human–environment
interactions in agricultural systems. Environmental Modelling & Software 26(7), 845–859
(2011)
[5] Zato, C., Villarrubia, G., Sánchez, A., Bajo, J., Corchado, J.M.: PANGEA: A New Plat-
form for Developing Virtual Organizations of Agents. International Journal of Artificial
IntelligenceTM 11(A13), 93–102 (2013)
[6] Mas, A.: Agentes software y sistemas multiagente: conceptos, arquitecturas y aplicaciones.
Prentice Hall (2005)
[7] Boella, G., Hulstijn, J., Van Der Torre, L.: Virtual organizations as normative multiagent
systems. In: Proceedings of the 38th Annual Hawaii International Conference on System
Sciences, HICSS 2005, pp. 192c–192c. IEEE (January 2005)
[8] Rodriguez, S., Julián, V., Bajo, J., Carrascosa, C., Botti, V., Corchado, J.M.: Agent-based
virtual organization architecture. Engineering Applications of Artificial Intelligence 24(5),
895–910 (2011)
[9] Julian, V., Rebollo, M., Argente, E., Botti, V., Carrascosa, C., Giret, A.: Using THOMAS for
Service Oriented Open MAS. In: Kowalczyk, R., Vo, Q.B., Maamar, Z., Huhns, M. (eds.)
SOCASE 2009. LNCS, vol. 5907, pp. 56–70. Springer, Heidelberg (2009)