Automation in Construction 12 (2002) 201 – 211

www.elsevier.com/locate/autcon

A framework for developing a unified B2B e-trading

construction marketplace
Heng Li a,*, Jiannong Cao b, Daniel Castro-Lacouture c, MiroslAaw Skibniewski d
a
Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China
b
Department of Computing, The Hong Kong Polytechnic University, Hong Kong, China
c
School of Civil Engineering, Purdue University, West Lafayette, IN 47907-1294, USA
d
Division of Construction Engineering and Management, School of Civil Engineering, Purdue University, West Lafayette, IN 47907-1294, USA

Abstract

With its great explosion and the advancement of related technology and services, such as the World Wide Web (WWW), the
Internet has provided a rich environment for developing Internet-based electronic commerce (e-commerce) applications in
construction. Among the different types of e-commerce, business-to-business (B2B) is the most widely used. The B2B e-trading
marketplaces (also called B2B-exchanges), which have sparked a revolution in the way of trading products between buyers and
suppliers, are an essential component of B2B applications. However, many e-trading marketplaces have been developed and
hosted by different construction products and services providers, and each forms a closed system with their own customers and
clients. In this paper, the concept of e-union is presented, which integrates the services provided by different e-trading
construction sites to provide an open e-trading service. The design of a mobile agent-enabled framework for building such an
open e-trading marketplace environment is described, along with a prototypical implementation.
D 2002 Published by Elsevier Science B.V.

Keywords: Framework; B2B; E-trading; Marketplace

1. Introduction gaged in business-to-business (B2B) electronic com-

Over the last few years, the Internet has evolved
from being a scientific network only, to a platform that
is enabling a new generation of business. More and
more companies and organizations are doing different
types of business and offer value-added services on the
Internet [1]. While the first stage was fueled by the
vision and innovation of business-to-consumer (B2C)
Internet companies, the current phase is defined by
the leadership and market success of companies en-
* Corresponding author. Tel.: +852-2766-5111; fax: +852-276-
42572.
E-mail address: bshengli@polyu.edu.hk (H. Li).
merce (e-commerce) [2].
The B2B e-trading marketplaces, which allow
large communities of buyers and suppliers to meet
and trade with each other, are an essential component
of B2B e-commerce applications. They resemble
stock exchanges in many ways, including the way
they are set up and organized and the trading methods
they employ—but they are trading physical commod-
ities such as doors, tiles and steel. They enable a
many-to-many relationship between multiple buyers
and sellers in the construction industry, who come
together and find each other in the cyberspace. They
allow participants to access various mechanisms to
buy and sell almost anything, from services to direct

0926-5805/02/$ - see front matter D 2002 Published by Elsevier Science B.V.

PII: S 0 9 2 6 - 5 8 0 5 ( 0 2 ) 0 0 0 7 6 - 6
202 H. Li et al. / Automation in Construction 12 (2002) 201–211
materials. Buyers and suppliers leverage economies
of scale in their trading relationships and access a
more ‘‘liquid’’ marketplace. Sellers find buyers for
their goods, buyers find suppliers with goods to sell.
Many-to-many liquidity allows the use of dynamic
pricing models, thus further improving the economic
efficiency of the market.
Construction companies are now conducting their
business using Web-based e-commerce system. Many
believe that e-commerce can provide a win-win sit-
uation for both suppliers and buyers, as e-commerce
can provide an expanded marketplace within which
buyers and suppliers can communicate directly with
each other. E-commerce might bring the answer
awaited for clients or construction firms, that is, to
create the solution for the procurement of materials
using non-traditional methods, avoiding delays, high
prices, lack of specified products, etc. [3]. Online
construction trading markets are not limited by the
physical limitations of store spaces and can carry a
much larger variety of products and different styles and
sizes. At the same time, buyers can search through a
wide range of products with low transaction costs at
any time convenient to them. More importantly, the
direct communication between buyers and suppliers
will cut off the multiple layers of middlemen between
suppliers and buyers. These middlemen take commis-
sions and fees from both buyers and suppliers. The use
of e-commerce will therefore directly benefit the
buyers so they can efficiently purchase cheaper prod-
ucts with a variety of choices [4].
The last several years have witnessed the emergence
of online B2B e-trading marketplaces. Well-known
examples include Catex, Chemdex, e-STEEL, Metal-
Site, and VHCome [5]. Currently, however, it is often
the case that, within a particular industry such as the
construction industry, many e-trading marketplaces
have been developed, owned and/or hosted by different
companies. Each of the e-trading marketplaces forms a
closed system with their own customers and clients.
The totality of these e-trading marketplaces appears to
be islands in the sea, as they are isolated and with no
interoperation between each other. In this paper, a
framework is presented for developing an interoperable
e-trading marketplace for the construction industry by
linking all the existing e-trading marketplaces, as any
single e-trading marketplace may not be able to meet all
the requirements of the buyers. The ability of market-
places to interoperate extends the idea of liquidity and
network effect by joining more buyers with more
suppliers but does not sacrifice the ability of each
marketplace to be highly specific to the supply-chain
node or target buyer group it serves. The concept of
‘‘e-union’’ is described, which integrates the services
provided by different e-trading marketplaces in the
construction industry to provide an open and unified
e-trading marketplace, or e-union, which is enabled by
the use of a mobile agent.
2. The e-union concept
The unique feature of a B2B e-trading marketplace
is that it brings multiple buyers and sellers together (in a
‘‘virtual’’ sense) in one central market space and en-
ables them to buy and sell from each other at a dynamic
price which is determined in accordance with the rules
of the exchange. On the Internet, every web site, which
enables buyers and sellers to come together and find
each other, is really a ‘‘virtual’’ e-trading marketplace.
An e-trading marketplace was developed as part of
VHBuild, a Web-based mediation service for construc-
tion project management in Hong Kong [6]. Fig. 1
shows selected screen dumps of the system’s web
interface. The service is provided to mediate among
the different parties involved in a construction project
for various project management activities, including
project initiation, tendering, materials purchasing,
project monitoring, project information querying,
decision making, etc. These activities are systemati-
cally organized and coordinated around a construction
project, with a close modeling of the data flow and
workflow in the building and construction industry.
The system also provides an e-trading marketplace for
advertising, marketing and buying/selling of construc-
tion materials and equipment.
The Web-based e-trading marketplace contains two
major functions: providing trading information and
facilitating trading transactions. Users will log onto
the system as either buyers or sellers of construction
materials. The sellers can upload their product infor-
mation and find out summary information about their
customers and their transactions. The buyers can search
certain types of products, giving their requirements
such as brand, model, quality, price, etc., or they can
browse the products on display. When the products
 H. Li et al. / Automation in Construction 12 (2002) 201–211 203

Fig. 1. VHBuild, a web-based mediation service for construction project management.

requested by a buyer are found, the information will be
displayed to the buyer and the contact information of an
agent closest to the buyer’s location will also be sent to
the buyer. The buyer can negotiate with the agent about
the price, or send an order to the agent by filling some
forms online. If the user could not find the products he/
she wants, the system will keep their inquiry in a place,
which can be visited by the sellers.
In recent years, several other construction e-trading
sites have appeared on the Web. They are operated by
different organizations and attracted different group of
clients. They specialize in trading materials from
suppliers in different regions. From time to time,
however, clients of one e-trading system would like
to buy some materials that are not available in that
system but offered in other sites. One way to solve
this problem is to let the client register at several
e-trading marketplaces and search these sites one by
one. Fig. 2 illustrates this approach.
To provide better value-added services to the
clients, we present the concept of an e-trading union
by linking together relevant e-trading marketplaces so
that cooperation between the e-trading marketplaces
can be facilitated. In the e-trading union, when an
e-trading site does not have the material requested by
one of its registered clients, the site will send a request

Fig. 2. Traditional approach to searching and buying a product.

204 H. Li et al. / Automation in Construction 12 (2002) 201–211
Fig. 3. Open e-trading union.
to other e-trading sites in the union. The remote
e-trading site will treat the request as one from an
associated client. An associate client cannot directly
make use of the service of the e-trading marketplace
as its registered client. When the material requested is
found, the remote site will inform the requesting
e-trading site, which will decide whether to put an
order on behalf of its client. Once a deal is made
between the two e-trading sites, the requesting site
will inform the remote site the contact information of
an agent to whom the material should be delivered; it
will also display the material information to the buyer,
together with the contact information of an agent
closest to the buyer’s location. The remote e-trading
site will charge some extra amount of money, usually
a certain percentage of the normal price, for providing
the service. Fig. 3 illustrates this open marketplace
approach.
3. A mobile agent-enabled framework for e-union
Since individual e-trading systems are developed
based on different computational architectures, plat-
forms and software, it is not a trivial task to make them
interoperable. There are several issues that need to be
addressed for realizing the e-union concept. The first
issue is how an individual e-trading system finds the
service from other sites and communicates with each
other. The second issue is how one system understands
the various kinds of data from other systems, such as
the query, order and product information. Particularly,
the date may be in different formats and databases for
the representation and storage. In order to solve these
problems, we developed a mobile agent-based frame-
work for implementing the e-union.
A mobile agent is a computer program that can
autonomously migrate between network sites, i.e. it can
execute at a host for a while, halts execution, dispatches
itself (together with its data and execution state) to
another host and resumes execution there—all under its
own control [7,8]. It has been found that mobile agent is
especially suitable for structuring and coordinating
distributed applications running in a wide-area envi-
ronment like the Internet [9 –11]. Such an environment
is characterized by a larger number of heterogeneous
nodes with dynamically changing services and resour-
ces, high variation of connectivity both in performance
and in reliability, high variation of workload and net-
work traffic, mobility of hosts, etc. Mobile agents can
move through the network of sites to search for, filter
and process information they need to accomplish their
tasks [12 –14]. Sending mobile agents to the remote
server results in a large performance improvement,
because they can reduce the number of times one site
contacts another and they can filter out non-useful
information and thus reduce the consumption of com-
munication bandwidth. Furthermore, mobile agent
 H. Li et al. / Automation in Construction 12 (2002) 201–211
brings flexibility and scalability into distributed,
dynamic systems due to its ability to encapsulate
policies, convention and algorithms and its ability to
be dynamically created and destroyed.
In our framework, an e-trading site joints the union
by broadcasting its address and service interface to
other members in the union. The membership infor-
mation is kept at each e-trading site in a database. Mo-
bile agents are dispatched to other member sites for
querying and ordering construction materials that are
not available at the local site. Fig. 4 shows the ar-
chitecture of the mobile agent-enabled e-union system.
The workflow of the e-union can be described as
follows. The client first issues a query for information
about a certain construction material. The local e-tra-
ding site searches its material database and finds out
that the requested material is not available. It will then
forward the client query to the Query Generator. The
Query Generator looks up the e-union database for
Fig. 4. Mobile agent-enabled e-union architecture.
205
information about members in the union. It then gene-
rates remote queries by formatting tuples of e-trading
site address and query statements, and passes these
queries, together with search criteria, to the Static
Agent, which is the master agent. The Static Agent is
responsible of creating mobile agents to carry out the
remote queries. The created mobile agents will be
dispatched to remote e-trading sites. Upon arrival, a
mobile agent interacts with the remote e-trading site to
connect to the materials database server, executes the
query statements, analyzes and filters the retrieved
records, extracts useful data as result and travels back
to the original e-trading site. After all mobile agents
return, the Static Agent at the original site will gather
the partial results and return the final result to the
client.
There are two types of mobile agents: Porter agents
and Traveler agents. A Porter agent is dispatched to a
particular remote e-trading site, executes the given
206 H. Li et al. / Automation in Construction 12 (2002) 201–211
Fig. 5. Interaction diagram of an e-business/e-trading solution for
construction materials management using UML.
query, retrieves result and then sends back the result to
the Static Agent at the original site. A traveler agent,
on the other hand, can travel with an itinerary over the
Internet, from one e-trading site to another. At each
visited site, it functions in the similar way as the
Porter agent. However, it can store the partial results
obtained at each site and carry them with it until it
finds the expected information or travels through all
the member sites in the e-union.
To facilitate the interoperability between the mate-
rial databases at each e-trading sites, we have devel-
oped a solution using XML. Databases and XML offer
complementary functionality for storing material data.
Databases store data for efficient retrieval, whereas
Fig. 6. An e-business solution for the procurement of steel reinforcement using XML [16].
XML offers an easy information exchange that enables
interoperability between different material databases.
In our solution, database tables are converted into XML
documents. These XML documents can then be used as
a data-exchange format, presented as HTML pages,
and searched with XML-based query languages. In this
way, the material records from different databases can
be combined in their XML document format and
analyzed and filtered by the mobile agents.
Another possibility for triggering actions on e-trad-
ing sites and conveying information to the registered
clients is by modeling XML applications through
unified modeling language, or UML. UML defines a
standard language and graphical notation for creating
models of business and decision-based systems [15].
The supply chain of construction materials is a suitable
environment for this endeavor, since there are per-
manent interactions among customer, supplier, other
e-sites and delivery agents. With the use of UML
diagrams featuring the workflow of information, it is
possible to generate the interactions and decision
processes necessary to reflect the abovementioned in-
teractions. Fig. 5 displays the workflow and interaction
diagram of typical construction materials e-business
applications.
The Internet serves as the environment where the
e-business system will provide an efficient coordina-
tion and communication of tasks and resources in the
procurement of construction materials. By facilitating
communication as a basis for information exchange
and conflict resolution, the proposed e-business model
 H. Li et al. / Automation in Construction 12 (2002) 201–211 207

Fig. 7. XML-enhanced e-union architecture.

is capable of providing the overall process of estima-
tion, revision and procurement of steel reinforcement
with tools for integration and cooperation along the
supply chain. The tools that constitute the architecture
of the collaborative e-business system, i.e. HTML
(Hyper Text Markup Language), ASP (Active Server
Pages) and XML, will allow substantial communica-
tion among participants through the Internet [16].
A proposed model of e-business interaction in
construction materials management using XML offers
a dynamic approach to the flow of information along
the model network, as seen in Fig. 6.
The bcXML was adopted as the data infrastructure
[17]. The bcXML stands for Building-Construction
XML developed by Information Societies Technol-
ogy, aimed at helping the European building and
construction industry to develop, demonstrate and
disseminate a new communication technology tailored
to the needs of construction industry. The bcXML is
the core component of the project eConstruct, which
addresses the problems of insufficient information
structuring of data exchange. The bcXML is seman-
tically rich and supports many of the notions used in
practice. In particular, the semantics included in
bcXML support the e-commerce communication
about construction products.
Fig. 7 describes the XML-enhanced architecture in
which a layer is placed between mobile agent and

Fig. 8. Major components of the prototype.

208 H. Li et al. / Automation in Construction 12 (2002) 201–211
Fig. 9. Screen dump of user interface for query.
material database. Instead of querying the database
directly, the mobile agent submits the query to the
XML Tool and gets back the result in XML document
format with the bcXML Schema. The XML Tool
receives the query request from the mobile agent,
requests the Parser to retrieve the records in XML
document. Because of the material database’s highly
regular data storage structure, the material data can be
first mapped into data-centric XML documents. The
Parser extracts the desired data from the database by
executing the query statements. The data in table
format is then transformed into an XML document
with the bcXML Schema.
In order for the Parser to generate XML documents
validated with bcXML Schema, the material database
should be constructed following the specification of
Fig. 10. Search results by Porter agents.
the bcXML. The material name, data type, data unit
should be stored by using bcDictionary [17] and the
database schema should match the bcXML Schema.
4. Prototypical implementation and experiments
To demonstrate the effectiveness of the proposed
framework, we have implemented a prototype of the
e-union using the IBM Aglet mobile agent platform
[18]. The Aglet is a Java-based mobile agent frame-
work. The IBM Aglet platform also provides an aglet
viewer called Tahiti and aglet servers, which are
powerful machines that can host large number of
aglets, Mobile agents in our framework are imple-
mented as aglets and use the ATP protocol for
interactions. Using ATP, the Aglet Tahit Server
installed at each e-trading site allows the mobile
agents to navigate and perform query on the materials
data. Oracle 8 is used as the materials database server.
In this prototypical implementation, we did not imple-
ment the XML-enhanced architecture.
Fig. 8 shows the major components of the proto-
type. The Query Generator is implemented with a Java
servlet program, which accepts client’s request via the
user interface (see Fig. 9) and generates SQL state-
ments used for querying materials databases of remote
 H. Li et al. / Automation in Construction 12 (2002) 201–211 209

Fig. 11. Search results returned by Traveler agent.

e-trading sites. It is also responsible of generating the The method for connecting to the materials database at
itinerary for mobile agents to be dispatched. a visited e-trading site is similar to that used by a Porter
Both Porter agents and Traveler agents have been agent. After retrieving the material data from the
implemented. Porter agents are dispatched when paral- database, the Traveler agent stores the results into its
lel search is used. If a Porter agent arrives successfully ‘‘Pocket’’ which are carried by the agent throughout
at the destination e-trading site, it extracts the SQL the whole itinerary. In each site, when the desired
statements attached in the message carried by it. Before material records are found, they will be compared with
executing the SQL statements, the Porter agent first the records which were found in the previous sites in
gets the information from the Aglet Server at the visited order to filter out redundant or useless information.
site on connecting to the materials database at the site. New and useful records are added to the Pocket. After
The information includes the connection string, user-
name/password for connecting the database and the
database driver. The Porter agent then establishes a
connection to the database. After the connection is
established, the Porter agent gets the data from the
database. Before returning the results back to the
applet, the Porter constructs the data into HTML table
format. Fig. 10 shows the results returned by three
Porter agents. The master agent at the original e-trading
site needs to assemble the results from all the Porter
agents into the final result and sends to the client.
On the other hand, a Traveler agent can be dis-
patched to travel through all the specified remote sites. Fig. 12. Total time for searching within five e-trading sites.
210 H. Li et al. / Automation in Construction 12 (2002) 201–211
Fig. 13. Total time for searching at a single e-trading site.
the whole trip completes, the Traveler agent returns the
results to the master agent at the original e-trading site.
The Traveler agent can also be instructed to return
either the first-fit result or the best result (see Fig. 11).
Experiments have been performed to evaluate the
performance of the mobile agent-enabled framework.
Five SUN workstations are set up with Aglet Tahiti
servers and iPlanet Web servers supporting Java
Servlets. For comparison, we have also implemented
a message-passing version using Java servlets com-
munications. In this version, the Query Generator at
the local e-trading site queries to the e-trading Union
Information database and forms a sequence of pairs of
address and query statement, one for each e-trading
Fig. 14. Network traffic generated by the Traveler and Servlet approaches.
site to search. It then submits the information to the
local Query Submit Servlet, which submits the query
to the remote e-trading sites one by one. The Query
Execute Servlet at a remote site receives the query,
executes it and then returns the result.
Fig. 12 shows the average total time for the three
different approaches. It can be seen that the Porter
agent approach performs the best. This is because of
the parallelism in their executions. However, this
approach will result in many mobile agents being
dispatched in the network, increasing network traffic.
In addition, as we can observe from Fig. 13, in terms
of the time spent on each individual remote site,
Porters perform worse than the other two approaches.
From Fig. 13, we can see that the Traveler
approach is faster than the Servlets approach. How-
ever, it may not reflect the real situation. In our
experiments, both Tahiti server and iPlanet server
were shut down and restarted before collecting the
next set of statistics data. This is to prevent the effect
of caching of the previous result in memory on the
performance, especially in the servlet/web server.
When the servers were not shut down, we observed
that the time required by the two approaches became
closer. On some occasions, the performance of the
servlet approach was even better.
 H. Li et al. / Automation in Construction 12 (2002) 201–211
Fig. 14 compares the performance of the Travelers
approach with the Servlets approach in terms of the
volume of data records transferred on the network. It
shows that the Traveler approach significantly reduce
network traffic.
5. Conclusion and future work
B2B e-trading marketplaces for the construction
industry present ideal structures for commercial
exchange, achieving new levels of market efficiency
by tightening and automating the relationship between
suppliers and buyers. Based on their ability to bring
buyers and sellers together online and thereby to create
dynamic pricing, B2B e-trading marketplaces are the
killer application in the B2B Internet revolution. In this
paper, we have presented the concept of e-union, which
integrates the services of separate e-trading market-
places to provide open, value-added e-trading services.
We have described a mobile agent-enabled framework
for building the e-union. A prototypical implementa-
tion was also presented which demonstrates the feasi-
bility of the proposed e-union concept. Experiment
results of performance evaluation have been reported.
Our future work includes implementing the XML-
enhanced architecture to achieve true interoperability
of the e-trading sites. Further experiments to evaluate
the various design tradeoffs and performance are
planned. Finally, we will investigate alternative ap-
proaches to building e-union, such as using Common
Object Request Broker Architecture (CORBA) or
integrating mobile agents with CORBA.
References
[1] D.L. Hoffman, T.P. Novak, P. Chatterjee, Commercial scenar-
ios for the Web: opportunities and challenges, Journal of Com-
puter-Mediated Communication 3 (3) (1995).
[2] B. Sculley, W. William, B2B exchanges: the killer application
in the business-to-business internet revolution, Proc. 31st Ha-
waii Int’l Conf. On System Sciences (HICSS31), Kona, Ha-
waii, Jan. 2000, University of Hawaii, Hawaii, pp. 149 – 154.
[3] D. Castro-Lacouture, D. Harmelink, M. Skibniewski, Appli-
cability of CAD software enabled with artificial intelligence for
drawing management and rebar take-off practices in the US.
211
Unpublished report prepared for VHSoft Software, Hong
Kong, Division of Construction Engineering and Management,
School of Civil Engineering, Purdue University, May 2001.
[4] J.Y. Bakos, A strategic analysis of electronic marketplaces,
MIS Quarterly 15 (3) (1991) 295 – 310.
[5] S. Kong, H. Li, P.E.D. Love, An e-commerce system for
construction material procurement, Construction Innovation
1 (1) (2001) 43 – 54.
[6] H. Li, T.C. Wong, Intelligent web-based project management
system for large scale construction projects. Beijing, Proceed-
ings of Construction Management of Large-Scale Projects,
13 – 16 December 2000, Construction Engineering Press, Bei-
jing, pp. 149 – 152.
[7] D. Chess, C. Harrison, A. Kershenbaum, Mobile agents: are
they a good idea? IBM Research Report, RC 19887 (#88465)
3/16/95, 1995.
[8] D.B. Lange, M. Oshima, Seven good reasons for mobile
agents, Communication of the ACM 42 (3) (1999) 88 – 89.
[9] J. Cao, G. Chan, W. Jia, T. Dillion, Checkpointing and rollback
of wide-area distributed applications using mobile agents, Proc.
IPDPS 2001, University of San Francisco, San Francisco, USA,
April, 2001, pp. 22 – 34.
[10] S. Funfrocken, Integrating Java-based mobile agents into web
servers under security concerns, Proc. 31st Hawaii Int’l Conf.
On System Sciences (HICSS31), Kona, Hawaii, Jan. 1998,
University of Hawaii, Hawaii, pp. 34 – 43.
[11] R.S. Gray, Mobile agents: motivations and state-of-the-art sys-
tems, Tech. Report, Dept. of Computer Science, Dartmouth
College, April 2000. http://agent.cs.dartmouth.edu/.
[12] P. Dasgupta, N. Narasimhan, L.E. Moser, P.M. Melliar-Smith,
MAgNET: mobile agents for networked electronic trading,
IEEE Transactions on Knowledge and Data Engineering 11
(4) (July – August 1999) 509 – 525.
[13] S. Papastavrou, G. Samaras, E. Pitoura, Mobile agents for
world wide web distributed database access, IEEE Transac-
tions on Knowledge and Data Engineering 12 (5) (Septem-
ber – October 2000) 802 – 820.
[14] T. Sandholm, Q. Huai, Nomad: mobile agent system for an
internet-based auction house, IEEE Internet Computing 5 (2)
(2000) 80 – 86.
[15] D. Carlson, Modeling XML Applications with UML, Addison
Wesley, New York, 2001.
[16] D. Castro-Lacouture, M. Skibniewski, Development of an
e-business solution for the integration of steel reinforce-
ment supply chain in construction projects, Proceedings
for the First International Conference on Construction the
21st Century (CITC2002), Challenges and Opportunities in
Management and Technology. 25 – 26 April, 2002, Florida
International University, Miami, 2002, pp. 197 – 204.
[17] M. Bohms, G. te Brake, P. Bonsma, F. Tolman, R. van Rees,
Draft version of the bcXML Specification (D103draft),
D103-bcXMLdraft, IST-1999-10303 eConstruct, http://www.
eConstruct.com, 1999.
[18] D.B. Lange, M. Oshima, Programming and Deploying Java
Mobile Agents, With Aglets, Addison Wesley, New York, 1998.