Automation in Construction 14 (2005) 413 430 www.elsevier.

com/locate/autcon

An agent-based framework for supply chain coordination in construction

Xiaolong Xuea, Xiaodong Lia, Qiping Shenb,*, Yaowu Wanga
b a School of Management, Harbin Institute of Technology, PR China Department of Building and Real Estate, The Hong Kong Polytechnic University, Hong Kong

Received 1 February 2004; received in revised form 1 July 2004; accepted 1 August 2004

Abstract Supply chain coordination has become a critical success factor for supply chain management (SCM) and effectively improving the performance of organizations in various industries. Coordination refers to the integration of different parts of an organization or different organizations in supply chain to accomplish a collective set of tasks and to achieve mutual benefits. This paper defines the concepts of construction supply chain (CSC) and construction supply chain management, especially regards construction supply chain management as the coordination of interorganizations decision making in construction supply chain and the integration of key construction business processes and key members involved in construction supply chain. Much research and practice indicate that there still are many problems in construction, most of which are supply chain problems. The research analyzes the problems in construction supply chain. In order to resolve these problems and improving the performance of construction, an agent-based framework for construction supply chain coordination is designed based on the agent technology and multiattribute negotiation and multiattribute utility theory (MAUT). The framework, which integrates the construction organizations in construction supply chain and multiattribute negotiation model into a multiagent system (MAS), provides a solution for supply chain coordination in construction through multiattribute negotiation mechanism on the Internet. Finally, the prototype of the framework is developed and tentatively run based on an imaginary construction project. The trial run reveals the feasibility to implement the agent-based framework for coordination in construction. D 2004 Elsevier B.V. All rights reserved.
Keywords: Construction; Coordination and management; Intelligent agent; Multiattribute negotiation; Multiagent systems; Supply chain

1. Introduction In recent years, the application of supply chain management (SCM) philosophy to the construction
* Corresponding author. E-mail address: bsqpshen@polyu.edu.hk (Q. Shen). 0926-5805/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.autcon.2004.08.010

industry has been widely investigated as an effective and efficient management measure and strategy to improving the performance of construction, which has suffered from high fragmentation, large waste, poor productivity, cost and time overruns, and conflicts and disputes for a long time [13], and to address adversarial interorganizational relationship of organ-

414

X. Xue et al. / Automation in Construction 14 (2005) 413430

ization by increasing number of construction organizations and researchers [1,412]. SCM can be considered as the coordination of distributed decision making of organizations or participants on material flow, information flow, human flow, and cash flow in supply chain from systems perspective. According to Lau et al. (2004) [48], SCM is defined as bcoordination of independent enterprises in order to improve the performance of the whole supply chain by considering their individual needsQ. This definition describes the main function and principle of SCM, i.e., coordination. Swaminathan and Tayur [13] classify SCM issues into two broad categories: configuration (design-oriented) issue that relates to the basic infrastructure on which the supply chain executes, and coordination (execution-oriented) issues that relate to the actual execution of supply chain. Schneeweiss and Zimmer [14] also regard SCM as a management activity that has to do with the coordination of logistic process being locally controlled by various independent organizations (decision-making units) in the environment of internationalization and globalization of markets together with an increased focus on organizations core competence. Up to now, supply chain coordination has become a very popular research topic among the research community in SCM and a vital management issue of organizations in the collaborativecompetitive business environment. Coordination is bmanaging the dependencies between activitiesQ [15]. It is defined as a mutually beneficial and well-defined relationship entered into by two or more organizations to achieve common goals. It also refers to the integration of different parts of an organization or different organizations in supply chain to accomplish a collective set of tasks and to achieve mutual benefits. It involves more formal relationships, objectives and actions which are mutual, compatible and common, not necessary a centralized authority [16]. Multiagent systems (MAS) technology offers new means and tools for supply chain coordination [17,18]. According to Wooldridge and Jennings (1995) [49], an agent is a self-contained program capable of controlling its own decision making and acting based on its perception of its environment, in order to one or more goals. An agent must possess any two of the following three behavioral attributes: autonomy, cooperation, and learning [19]. MAS comprises a number of intelligent

agents, which represents the real world parties and cooperate to reach the desired objectives. In MAS, each agent attempts to maximize its own utility while cooperating with other agents to achieve their goals [20]. The main advantage of MAS is its responsibilities for acting various components of the engineering process or participants of the business process which is delegated to a number of agents. MAS is suitable for domains that involve interactions between different organizations with different objectives and proprietary information [21]. Based on these discussions, we can clearly see that SCM system is a typical MAS, where the participants are delegated to different agents. Furthermore, agent-based supply chain coordination has been proved to be an effective mechanism to improve the performance of SCM [2224]. The core principles of SCM and agent technology provide new perspectives for construction supply chain (CSC) management. However, little research has been conducted to investigate the application of intelligent agent to coordination problems in CSC. The Centre for Integrated Facility Engineering of Stanford University established a distribution cooperative CAD environment entitled AgentCAD, which presented a framework for collaborative distributed facility engineering [25]. Anumba et al. [26] presented the key features of an agent-based system for collaborative design of portal frame structures and made a significant contribution by allowing for peer to peer negotiation between design agents. Pena-Mora and Wang [27] proposed a collaborative negotiation methodology and a computer agent named CONVINCER, which incorporates that methodology to facilitate or mediate the negotiation of conflicts in large-scale civil engineering projects. Min and Bjornsson [28] presented a conceptual model of agent-based supply chain automation, in which a project agent gathers actual construction progress information and sends to subcontractor agents and supplier agents, respectively, over the Internet. They evaluated an agent-based SCM model compared with traditional SCM practice through simulation. Ren et al. [29] developed multiagent system for construction claims negotiation (MASCOT) to resolve inefficiency problems. Kim and Paulson [30] presented an agentbased compensatory negotiation methodology to facilitate the distributed coordination of project schedule changes wherein a project can be rescheduled dynamically through negotiation by all of the concerned

X. Xue et al. / Automation in Construction 14 (2005) 413430

415

subcontractors. The methodology consists of a compensatory negotiation strategy based on utility of timing, multilinked negotiation protocols, and message-handling mechanisms. Whereas these researchers have addressed a number of key issues in the field of CSC, very few researchers have considered CSC management as the coordination of interorganizations decision making in CSC and apply the agent technology into the whole CSC management from a systemic perspective. Most research projects tend to look into agent-based internal coordination of an organization, which focuses on one of the stages in CSC, such as design coordination, project schedule changes coordination, and construction claims and conflicts resolving. These approaches may lead to the optimization of a number of subsystems; however, the sum of the optimized subsystems may present a problem to the optimization of the system as a whole. This paper presents an agent-based framework for supply chain coordination in construction (ABS3C) based on multiattribute negotiation and utility theory, which integrates design and construction, and organizations (or participants), including owner, designer, general contractor (GC), subcontractors, and suppliers in CSC. This integrated coordination framework extends the internal supply chain of general contractor to external supply chain of designer, subcontractors, and suppliers. In the decision-making process of participants in CSC, the factors: cost, time, quality, safety, and environment normally are considered as the main decision-making variables. Here, we regard the five factors as five attributes of decision making in the construction process. An agent-based multiattribute negotiation mechanism is designed for supporting the realization of the framework. The research methodology is described. The concepts and problems of CSC are defined and discussed. A prototype system of the agent-based framework for supply chain coordination in construction has been evaluated through the use of a hypothetical construction project and a toolkit called ZEUS.

2. Theoretical foundation According to Pena-Mora and Wang [27], negotiation theory is the study of exchanges between

participants to reconcile their differences and produce a settlement. Negotiation is defined as one kind of decision-making process where two or more participants jointly search for a space of solution with the goal of achieving consensus [31]. From this definition, it can be found that negotiation is an effective mechanism for supply chain coordination in construction. A successful negotiation needs a number of elements in the negotiation process. For example, the participants have a commitment to settle the matter at hand; good communication skills are required; there is an agreement between participants on what is actually the matter under negotiation. There are three outcomes of negotiation: win/win, win/lose, and lose/ lose. The win/win outcome meets the needs of both participants [32]. One of the objectives of our research is to achieve the win/win outcome. Since a number of factors, such as cost, time, quality, safety, and environment, must be considered in the decision-making process of CSC management, our research adopts the multiattribute negotiation technology to coordinate the CSC. These factors are regarded as the attributes involved in CSC decision making. The multiattribute negotiation technology is developed based on the multiattribute utility theory (MAUT), which is an analytical tool for making decisions involving multiple interdependent objectives based on uncertainty and utility analyses [33] and an evaluation scheme for estimating various products and performance [34,35]. This research adopts MAUT to evaluate CSC decision-making problems. The procedure of MAUT application to supply chain coordination in construction is shown in Fig. 1 (adapted from Ref. [36]), where a decision is one of a set of possible solutions to the supply chain coordination problems in construction and an attribute is one of a set of factors that will influence the CSC decision-making process. The application of MAUT will be discussed in the subsequent sections in detail. According to Gupta et al. [17], agent technology provides a convenient tool to enhance communication and coordination in a distributed system to ensure efficient decision making. Kwon and Lee [22] also believe that MAS, where multiple agents work collectively to solve specific interorganization problems, provides an effective platform for coordination across organizations in the supply chain. This research adopts multiagent technology to improve the coordi-

416

X. Xue et al. / Automation in Construction 14 (2005) 413430

Fig. 1. Procedure of using MAUT in supply chain coordination in construction.

nation of CSC decision making. According to the definition of MAS, CSC management system is a typical MAS. In this kind of MAS, the partners, i.e., general contractor (GC), owner, designer, subcontractors, and suppliers are delegated to corresponding agents, i.e., GC agent, owner agent, designer agent, subcontractor agents, and supplier agents. These agents are designed and realized in ZEUS, an advanced developing environment for distributed multiagent systems and has been used to develop agent-based coordination systems in construction [26,29,37,38]. Multiagent technology is integrated to negotiation and multiattribute utility theory in the research. Each agent possessing own preferences and utilities autonomously negotiates to other agents. Thus, agents can act collectively and efficiently as a society and cooperate to achieve their own goals as well as the common goals of the whole CSC management.

3. Concepts and problems of CSC 3.1. The concepts of CSC CSC consists of all construction processes, from the initial demands by the client/owner, through design and construction, to maintenance, replacement and eventual demolition of the projects. It also consists of organizations involved in the construction process, such as client/owner, designer, GC, subcontractor, and suppliers. CSC is not only a chain of construction businesses with business-to-business relationships but also a network of multiple organizations and relationships, which includes the flow of information, the flow of materials, services or products, and the flow of funds between owner, designer, GC, subcontractors, and suppliers [8,9,39 41]. According to Muya et al. [42], there are three types of CSC: the primary supply chain, which delivers the materials that are incorporated into the

final construction products; the support chain, which provides equipment and materials that facilitate construction, and the human resource supply chain which involves the supply of labor. In this paper, CSC is considered from the stage of owner demands to the stages of design, construction, and handover. A typical model of CSC is shown in Fig. 2. In the model of CSC, GC is the core of the CSC. And the owner and designer are the other two main partners in CSC. Excepting the direct suppliers of GC, subcontractors are also regarded as the suppliers of GC; meanwhile, subcontractors have their own suppliers. Although a number of researchers have provided definitions for CSC management (e.g., Refs. [12,43]), for consistency, this research defines CSC management as follows: CSC management is the coordination of interorganizations decision making in CSC and the integration of key construction business processes and key members involved in CSC, including client/ owner, designer, GC, subcontractors, suppliers, etc. CSC management focuses on how firms utilize their suppliers processes, technology and capability to enhance competitive advantage. It is a management philosophy that extends traditional intra-enterprise activities by bringing partners together with the common goal of optimization and efficiency. CSC management emphasizes on long-term, win/win, and cooperative relationships between stakeholders in systemic perspective. Its ultimate goal is to improve construction performance and add client value at less cost. We have identified eight key construction business processes that are implemented within the CSC across organizational boundaries. They are: project management, client service management, supplier relationship management, demand management, order fulfillment, construction flow management, environment management, and research and development. Fig. 3 presents a schematic view of CSC management.

X. Xue et al. / Automation in Construction 14 (2005) 413430

417

Fig. 2. Model of construction supply chain.

3.2. Problems in CSC Although there have been many changes in the construction industry as a result of the development of technology and culture over the last decades, CSCs do not seem to have changed much. Many problems still exist in CSC. According to Ref. [40], the major

problems originate at the interfaces of different participants or stages involved in the CSC, as shown in Fig. 3. The problems are caused by myopic and independent control of the CSC. Love et al. [12] and Mohamed [44] noted the highly fragmented characteristics of the construction industry. For example, the separation of design and

Fig. 3. Problems in CSC (adapted from Ref. [40]).

418

X. Xue et al. / Automation in Construction 14 (2005) 413430

construction, lack of coordination and integration between various functional disciplines, poor communication, etc., are the important impact factors causing performance-related problems, such as low productivity, cost and time overrun, conflicts, and disputes. Palaneeswaran et al. [11] revealed the weak links in CSC as follows: ! ! ! ! ! Adversarial relationships between clients and contractors; Inadequate recognition of the sharing of risks and benefits; Fragmented approaches; Narrow minded bwin/loseQ attitudes and short-term focus; Power domination and frequent contractual noncommitments resulting in adverse performance track records with poor quality, conflicts, disputes, and claims; Prime focus on bid prices (with inadequate focus on life-cycle costs and ultimate value); Less transparency coupled with inadequate information exchanges and limited communications; Minimal or no direct interactions that foster sustainable long-term relationships.

! ! !

In order to overcome the shortcomings (weak links) of CSC and resolve the problems in CSC, and to further improve the performance of the whole CSC, this research presents a solution which integrates the agent technology and multiattribute negotiation technology. This solution will be explained in details in the following sections.

4. Design of agent-based framework for supply chain coordination 4.1. Integrated design CSC involves multiple agents that delegate organizations to autonomously perform tasks through exchanging information. It has increasing number of participants due to the increasing scale and complexity of construction projects. As a result, the coordination among the participants becomes a challenge that is vital to the performance of CSC and the value to the client. These

require an agent-based framework which has an appropriate structure and effective coordination mechanism to promote efficient communication among all parties. The framework should also be stable, flexible, and user-friendly. Based on these considerations, we have designed the framework as shown in Fig. 4. In this framework, the domain agents include both dserviceT agents (i.e., coordinator agent, monitor agent, and name server agent) and dspecialtyT agents (i.e., owner agent, design agent, GC agent, subcontractor agents, and supplier agents). The suppliers include both GCs suppliers and subcontractors suppliers. Here, we assume that all materials and human resources are arranged by GC or subcontractors and there are no owners suppliers in this framework. All agents communicate and cooperate through the Internet. All specialty agents advertise their abilities, knowledge, and preferences in the acquaintance database maintained by of the coordinator agent, and their addresses in the address book maintained by the name server agent. The specialty agents use the coordinator agent to identify agents with the required abilities, knowledge, and preferences. They also use the name server agent to determine the addresses of the identified agents. The coordinator agent has a mailbox and message handler for receiving and responding to queries from agents about the abilities and preferences of other agents, and an acquaintance database for storing the abilities and prefernces of the agents. It functions by periodically querying all the agents in the society about their abilities and preferences, and storing the returned information in its acquaintance database. The monitor agent is used to view, analyse societies of agents. It functions by querying other agents about its states and processes, and then collating and interpreting the replies to create an up-to-date model of the agents collective behavior. This model can be viewed from different perspectives through visualization tools. The name server agent provides a look-up service for agents addresses and has a mailbox and message handler, the component needed for receiving and responding to agents requests for the addresses of other agents. The specialty agents, which are independent and autonomous in making decisions, are designed to

X. Xue et al. / Automation in Construction 14 (2005) 413430

419

Fig. 4. An agent-based framework for supply chain coordination in construction.

delegate corresponding behaviors of CSC participants. The GC agent stands for GC to perform its responsibilities, including cost control, schedule, quality, safety, and environment performance of the project through coordination with subcontrctors, suppliers, owner, and designers. The subcontractor agents simulate the decision-making process of subcontractors to negotiate with GC and their suppliers. The owner agent tracks the real schedule, cost, and quality of the project and duly send the information of design changes (new demands) to the designer agents and the GC agent. The designer agents will respond to the demands and send information and design change drawings to the owner and the GC. The GC agent reschedules the project, makes new decisions, and sends the information to subcontractor agents, supplier agents, and owner agent. In the framework, all coordination processes are based on multiattribute negotiation (bM-NegotiationQ as shown in Fig. 4) between sepcialty agents. According to Andreas [45], agentbased automated negotiation provides an effective mechanism to coordinate the decision-making activities in supply chain. The multiattribute negotiation mechanism for coordinating the interorganizations decsion making in CSC is the significant components of ABS3C. The following section describes the

design process of the multiattribute negotiation mechanism in detail. 4.2. Multiattribute negotiation model Although Kim and Paulson [30] present an agentbased compensatory methodology to improve the coordination of project schedule changes, it only considers the utility of timing and cost and is not suitable for the coordination of decision making in CSC. We present an agent-based multiattribute negotiation model for supply chain coordination in construction, which creatively extends the general negotiation model for A/E/C [27] from SCM and utility theory perspectives and integrates the compositional multiattribute negotiation model [46], as shown in Fig. 5. The model consists of three elements: CSC participants, multiattribute negotiation process, and the outcome. Each of these elements plays a role in a generic coordination problem within the domain. All participants reveal their attributes and preferences in the negotiation process with due regard to their parent organizations and relationships to the issues. Here, each of the CSC participants is represented by a corresponding agent in the framework. The outcome will be the negotiation results determined by the

420

X. Xue et al. / Automation in Construction 14 (2005) 413430

Fig. 5. Multiattribute negotiation model for supply chain coordination in construction.

interactive process between participants in CSC. The multiattribute negotiation process is the interactive exchange and elicitation of participants preferences. The following section will focus on the negotiation process. 4.3. Multiattribute negotiation process According to Jonker and Treur [46], the multiattribute negotiation process includes the following five steps: (1) (2) (3) (4) Evaluation of the attributes of the initial solutions made by the participants; These evaluations are aggregated into overall utilities of these initial solutions; Provision of the target utility; Based on the target utility and the distribution of attributes, the values of the target attributes are determined, which lead to a new round of decision making;

(5)

For each of the target attributes, an attribute value is chosen that has an evaluation value as close as possible to the target evaluation value for the attribute.

In the agent-based multiattribute negotiation model, these five steps are simplified to three processes: attributes evaluation, utility determination, and attribute planning. 4.3.1. Attributes evaluation As shown in Fig. 5, many attributes are involved in the process of supply chain coordination in construction. Attributes evaluation is a process whereby the value of the attributes are evaluated, based on the preferences of the participants in the CSC. All attributes are classified into two categories: quantitative and qualitative. For the quantitative attributes, such as cost and time, the value of these attributes can be directly calculated according to certain principles. For the qualitative attributes, such as quality, safety,

X. Xue et al. / Automation in Construction 14 (2005) 413430

421

and environment, it is necessary to construct a scale to measure the levels of these attributes. In this paper, a scale, from 0 (worst) to 10 (best), serves as the measure of the evaluation. The most frequently used five levels are: 1worse, 3bad, 5normal, 7 good, and 9better. When a compromise is needed, 0, 2, 4, 6, 8, 10 are intermediate values between the two adjacent judgments. After evaluating the attributes in a table form, an evaluation matrix that includes m participants and n attributes is obtained: X xij mn where x ij is the value of the jth attribute evaluated by the ith participant in a decision-making process. Next, the matrix X is translated into a unitary normalization matrix of vector: Y yij mn where 0Vy ij V1, and ,s m X yij xij x2 ij
i1

where U BOW is the utility of the own decision making, and the concession step (CS) is determined by CS b1 l=UBOW UBOT =UBOW where U BOT is the utility of the other participants decision making. The factor (1l/U BOW) expresses CS will decrease to 0 if the U BOW approximates the minimal utility l and (U BOTU BOW) expresses the current utility gap. b stands for the negotiation speed. 4.3.3. Attribute planning The attribute planning process refers to target evaluation and configuration determination [46]. The target evaluation of jth attribute TEj is given by TEj 1 sBTEj sEBOT; j where BTEj is the basic target evaluation of jth attribute, which is determined in such away that Aw j BTEj =TU. E BOT, j is the jth attribute evaluation value of other agent. s stands for the configuration tolerance. BTEj is determined as the below format: BTEj EBOW; j aj =N TU UBOW where E BOW,j is the jth attribute evaluation value of decision agent, N is a normalization factor, which is the weighted sum of a s with the P relative importance factors being the weightings: N= w j a j,. a j =(1w j ) (1E BOW, j ). The configuration determination for the next decision making includes three steps. Firstly, attribute values are determined with an evaluation that is as close as possible to the target evaluation value.

4.3.2. Utility determination In this process, the target utility is determined. The utility of the ith participants decision making (U i ) is given by Ui
n X j1

wj yij

Where the w j is the weighting of the jth attributes: n n n X 0XX wj akj akj
j1

Where j1 wj 1 and a kj is the value of relative weightiness between kth attribute and jth attribute. The value of a kj is given by
8 ; if kth attribute is more important than jth attribute <1 akj 0:5 ; if kth attribute is important as jth attribute : 0 ; if kth attribute is less important than jth attribute

Pn

k1 j1

Jonker and Treur [46] presented the formula of target utility (TU) of participants decision making as follows: TU UBOW CS
Fig. 6. Multiattribute negotiation protocol.

422

X. Xue et al. / Automation in Construction 14 (2005) 413430

Fig. 7. Roles represented in ABS3C.

Secondly, a partial configuration (excepting the quantitative attributes, such as cost and time) is selected from the closest value of attribute. The final step is to reevaluate the quantitative attributes. 4.4. Multiattribute negotiation protocol Negotiation protocol controls the interactions among agents by constraining the way the agents interact [30]. It also specifies the kinds of deals that the agents can make, as well as the sequence of offers and counteroffers that are allowed [29]. In our framework, multiple attributes are considered in the process of negotiation between agents. For example, GC agent needs to negotiate with subcontractor agents, supplier agents, designer agents, and owner agent regarding different attributes of solutions, such as time, cost, safety, and quality, based on the overall utility. The negotiation protocol is named as multi-

attribute negotiation protocol in this paper, which is shown in Fig. 6. Each agent offers its current best solution that is also saved locally. Then, the solution is sent to other agents. The sender waits for a message from other agents. If the message is acceptance of the offer, it indicates that the sent solution is consistent with the other agents solution, and the negotiation is successful. If the message is NoMoreSolution, it means that the other agent has run out of solutions. If the same is true for the initiating agent as well, then the negotiation ends unsuccessfully; otherwise, the agent will continue to generate new solutions. If the message includes a changed solution from the other agent, this solution is checked for compatibility with any of the past solutions generated by this agent. If an intersection is found, it presents a mutually acceptable solution, which determines the negotiation successfully. Otherwise, this indicates

Fig. 8. Interactions of agents in ABS3C.

X. Xue et al. / Automation in Construction 14 (2005) 413430 Table 1 Summary of Interactions Collaboration 1 2a 2b 3a 3b 4a 4b Registration Resolve query List query Location response List response Answer agent name Inquire agent name Explanation Agents notify the NS of their presence A request for the network location of a named agent A request for all agents of a particular type The location of the agent previously in question The list of agents previously in question NS answers the inquiries about the agent name Agent inquires from NS about the other agent name it needs to communicate with them Information about an agents current abilities and preference Ask for information about recipients abilities and preference Asks all activity be forwarded A copy of ant message sent Asks for agents with particular abilities and preference A list of agents matching the desired criteria A agent inquire from facilitator about the other agents abilities and preference The facilitator answers the inquires from agents GC negotiates the relative construction solutions with Subcontractors Subcontractors send their decision making to GC GC sends its demands plan to suppliers Suppliers send their supply plan to GC The owner inquires from GC about the project information and sends demands change GC reports the project information to Owner and response the demands change GC requires for relative problems of the design drawings Table 1 (continued) Collaboration 16 20 Explain Put forward new demands Explanation

423

21

Offer design drawings

Designer explains the design drawings The owner puts forward his new demands, i.e., design change The designer offers drawings to meet the owners demands

that utility gap remains between the initiating agent and other agents. This loop can be repeated. The main advantage of this protocol is that it guarantees the discovery of the Pareto optimum [47].

5a

Ability and preference response Ability and preference request Information request Activity notification Find request

5. Implementation of ABS3C The prototype is developed by using the ZEUS agent building toolkit. ZEUS is an advanced development toolkit for constructing distributed multiagent applications. It defines a MAS approach and supports it with a visual environment for capturing user specification of agents that are used to generate Java source code of the agents. ZEUS toolkit proposes five main technologies to enable agents to coordinate at the knowledge level: information discovery, communication, ontology, coordination, and legacy software problems. ZEUS is a culmination of a careful synthesis of established agent to provide an integrated environment for the rapid software engineering of collaborative agent applications [37]. The implementation process consists of the following steps: role modelling, application design, and trial run. 5.1. Role modeling in ABS3C ZEUS agent building toolkit adopts role modeling to address the specification, analysis, design, implementation, and maintenance of agents. Role models formalize the definition of an agent role and provide a comprehensible means of analyzing the problem in question. The role models are grouped into domains. The domains provide a context that enables developers to compare their planned system with existing applications. Role models, which

5b

6 7 8

6 10

Find response Inquire

11 12

Answer Negotiate with Subcontractors Sent solution (decision making) Send demands information Response GC demands Inquire project information and send change Report and response change

13 14 15 16

17

18

Require

424

X. Xue et al. / Automation in Construction 14 (2005) 413430

Fig. 9. Schedule of HCP building.

describe the dynamic interaction between roles, are architectural patterns that depict the high-level similarities between related systems, i.e., the problems inherent to each domain, but not how they were solved. The role models of ABS3C are illustrated in Fig. 7. The interactions between these roles are shown in Fig. 8. Table 1 summarizes the collaboration relationships. 5.2. Application design This process is illustrated based on a hypothetical construction project: HCP building. The HCP involves the following participants: owner, designer, general contractor, groundwork subcontractor, civil and structure subcontractor, building services subcontractor, finishing works subcontractor, concrete supplier, and finishing materials supplier. The project schedule is shown in Fig. 9. 5.2.1. Ontology creation Ontology is a set of declarative knowledge representing every significant concept within a particular application domain. The significance of a concept is easily assessed, if meaningful interaction cannot occur between agents without both parties being aware of it, then the concept is significant and must be modeled. Ontology contains the key concepts within the specific application domain, the attributes of each concept, the types of each attribute, and any restrictions on the attributes. In ZEUS, an individual domain concept is described by using the term dfactT. ZEUS provides two kinds of fact: abstract

and entity. In ABS3C, all the concepts refer to the entity. The ontology created in ABS3C is shown in Fig. 10. 5.2.2. Agent creation Agent creation includes three steps: agent definition, agent organization, and agent coordination. Agent definition determines the planning parameters, task and initial resources allocation. Agent organization illustrates the relationship between the own agent and other agents and acquaintance abilities from other agents. Agent coordination defines the coordination protocols and strategies between the own agent and other agents. Multiattribute negotiation and multiattribute negotiation protocol as

Fig. 10. Ontology in ABS3C.

X. Xue et al. / Automation in Construction 14 (2005) 413430 Table 2 Agent created in the ABS3C Agent name Specialty agents GC agent Roles played CSC head (Negotiation initiator, Manager and constructor) Client CSC participant (Negotiation partner, supplier, designer) CSC participant (Negotiation partner, supplier, constructor) CSC participant (Negotiation partner, supplier, constructor, consumer) CSC participant (Negotiation partner, supplier, constructor) CSC participant (Negotiation partner, supplier, constructor, consumer) CSC participant (Negotiation partner, supplier) CSC participant (Negotiation partner, supplier) Agent name server Monitor (to view, analyse or debug societies of agents) Coordinator (receive and respond to queries from agents about the abilities and preferences of other agents)

425

subordinates of subcontractors. The relations between agents in ABS3C are shown in Fig. 11. 5.2.3. Service and specialty agents configuration The service agents, i.e., agent name server, monitor and coordinator, are named as dutility agentsT in ZEUS. The monitor agent is given the dvisualizersT title, and the coordinator agent is labeled as dfacilitatorsT. Specialty agents, such as GC agent, civil and structure subcontractor agent (as seen in Table 2) are called task agents. Service and specialty agents configuration (as shown in Fig. 12) determines their running parameters in the prototype of ABS3C, e.g., the name of agent, the IP address of host, time grain, the name of address file, external program connected to agent, etc. 5.2.4. External program ZEUS allows users to link an external java class (program) to executing ZEUS agent program. Once linked to the agent program, external program can utilize the agents public methods to query or modify the agents internal state. The internal event model in ZEUS provides a mechanism whereby all significant events occurring in the agent can be monitored, for example, planning events, resource, acquaintance database events, message events, execution events, and coordination engine events. Using the event model, an external program that is linked to ZEUS agent can monitor particular events in the agent and react to them. In the prototype of ABS3C, each

Owner agent Designer agent

Groundwork subcontractor agent Civil and structure subcontractor agent

Building services subcontractor agent Finishing works subcontractor agent

Concrete supplier agent Finishing material supplier agent Service agents ANS agent Monitor agent Construction coordinator agent

previously mentioned are integrated into the prototype of ABS3C as the coordination protocols and strategy in ZEUS. Table 2 lists the agents in the prototype of ABS3C. GC agent as the CSC head possesses the core position comparing with other agents. GC is the superior of subcontractors and suppliers. The owner supervises the activities of GC and designer. The relationships of GC and owner, subcontractor and subcontractor are cooperative partnering. In HCP case, the finishing material supplier and concrete supplier are considered as the direct supplier of corresponding subcontractors, so the suppliers are considered as the

Fig. 11. Relations of agents in ABS3C.

426

X. Xue et al. / Automation in Construction 14 (2005) 413430

Fig. 12. Configuration of service and specialty agents.

specialty agent is linked to an external program as shown in Fig. 12. Each external program provides a user interface (negotiation window) through which the

users, for example, GC, subcontractor, supplier, etc., input their preferences on a decision making in the process of CSC.

Fig. 13. A screen shot when runs the prototype system.

X. Xue et al. / Automation in Construction 14 (2005) 413430

427

Fig. 16. Dialogue for new goal of agents.

Fig. 14. Negotiation window (GCUI).

5.3. Trial run and discussions The commands of running the prototype of ABS3C can be generated through the dCode GeneratorT panel in ZEUS. Fig. 13 gives a screen shot of the prototype system. When running the prototype, the negotiation

windows will display on the screen waiting for the decision makers input the preferences and relative weightiness of attributes. In the HCP case, we only consider five negotiation attributes: cost, time, quality, safety, and environment, which present the decision makers different preferences and utility, as shown in Fig. 14. The above information will be sent to the coordinator agent and corresponding negotiation agent. Fig. 15 displays the structure of HCP CSC and the process of information exchange and communication between service agent and specialty agent. When a new round of negotiation starts, the new goal of agent can be created through the dEnter Goal DialogT (Fig. 16). The prototype provides tool (monitor agent) to dynamically monitor the traffic volume of each agent and interagent traffic volume, as shown in Figs. 17 and 18. User can analyze and evaluate the performance and activities of agents. From Figs. 17 and 18, it can be concluded that the running process is stable. Although the frequency of communication of construction coordinator is much higher than other agents, this is not a problem. The phenomenon rightly reflects the hub role of the coordinator agent in the framework.

6. Conclusions Many researches and practices have proved that supply chain coordination has become the crucial strategy for successful SCM and effectively improv-

Fig. 15. Information exchange of agents in ABS3C.

428

X. Xue et al. / Automation in Construction 14 (2005) 413430

Fig. 17. Traffic volume of each agent.

ing the performance of organizations in various industries. This paper defines that concepts of CSC and CSC management, especially regards CSC management as coordination of interorganizations decision making in CSC and the integration of key construction business processes and key members involved in CSC. The research finds that there still are many problems in CSC. In order to resolve the problems and improve the performance of construction, an agent-based framework for construction supply chain coordination is designed based on the agent technology and multiattribute negotiation and utility theory. The framework, which integrates the organizations in CSC into an agent system, provides a solution for the supply chain coordination in construction through the application of multiattribute negotiation mechanism. The trial run reveals the

feasibility to implement ABS3C. The development of prototype system provides an effective platform to simulate the coordination process in CSC. It is helpful to reduce the duration of interorganization decision making and enhance the quick response ability to meet the change of demands in competitive marketing environment. Since this research on agent-based supply chain coordination in construction is an initial effort, some issues still need to be addressed. For example, the multiattribute negotiation mechanism for supply chain coordination in construction needs to be improved, especially how agents can elicit users preferences and utility functions relevant to negotiation-related decision making. Real construction process involves various thousands of participants and activities, so how to use agents to efficiently simulate the behaviors

Fig. 18. Interagent traffic volume.

X. Xue et al. / Automation in Construction 14 (2005) 413430

429

of decision making is a real challenge. The developed prototype system also has some components needed to be improved. This reveals that relative tools and approaches needed to be further developed for the implementation of ABS3C.

[11]

[12]

Acknowledgement The work described in this paper was supported by the Research Grants Council of the Hong Kong Special Administrative Region, P.R. China.

[13] [14]

[15]

References
[1] G.H. Briscoe, A.R.J. Dainty, S.J. Millett, R.H. Neale, Client-led strategies for construction supply chain improvement, Construction Management & Economics 22 (2) (2004) 193 201. [2] N. Daywood, A. Akinsola, B. Hobbs, Development of automated communication of system for managing site information using internet technology, Automation in Construction 11 (5) (2002) 557 572. [3] C.T. Formoso, L. Soibelman, C.M. De Cesare, E.L. Isatto, Material waste in building industry: main causes and prevention, Journal of Construction Engineering and Management 128 (4) (2002) 316 325. [4] W.J. OBrien, Construction Supply-Chain Management: A Vision For Advanced Coordination, Costing, and Control, NSF Berkeley-Stanford Construction Research Workshop, Stanford, California, 1999. [5] R. Vrijhoef, L. Koskela, The four roles of supply chain management in construction, European Journal of Purchasing and Supply Management 34 (6) (2000) 169 178. [6] K.A. London, R. Kenley, An industrial organization economic supply chain approach for the construction industry: a review, Construction Management & Economics 19 (2001) 777 788. [7] R.J. Arbulu, I.D. Tommelein, Contributors to lead time in construction supply chains: case of pipe supports used in power plants, Proceedings of Winter Simulation Conference: Exploring New Frontiers, 2002, pp. 1745 1751. [8] M. Saad, M. Jones, P. James, A review of the progess towards the adoption of supply chain management (SCM) relationships in construction, European Journal of Purchasing and Supply Management 8 (2002) 173 183. [9] R. Vrijhoef, L. Koskela, H. Voordijk, Understanding construction supply chains: a multiple theoretical approach to inter-organizational relationships in construction, Proceedings of International Group of Lean Construction 11th Annual Conference, Virginia, USA, 2003. [10] M.M.B. Azambuja, C.T. Formoso, Guidelines for the improvement of design, procurement and installation of elevators using supply chain management concepts, Proceedings of Interna-

[16]

[17]

[18]

[19] [20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

tional Group of Lean Construction 11th Annual Conference, Virginia, USA, 2003. E. Palaneeswaran, M. Kumaraswamy, S.T. NG, Formulating a framework for relationally integrated construction supply chains, Journal of Construction Research 4 (2) (2003) 189 205. P.E.D. Love, Z. Irani, D.F. Edwards, A seamless supply chain model for construction, Supply Chain Management: An International Journal 9 (1) (2004) 43 56. J.M. Swaminathan, S.R. Tayur, Modles for supply chains in Ebusiness, Management Science 49 (10) (2003) 1387 1406. C. Schneeweiss, K. Zimmer, Hierarchical coordination mechanisms within the supply chain, European Journal of Operational Research 153 (2004) 687 703. T.W. Malone, K. Crowston, The interdisciplinary study of coordination, ACM Computing Surveys 26 (1) (1994) 87 119. C.Y. Wong, J. Johansen, H.H. Hvolby. Supply chain coordination problems: literature review. Working Paper, No. 08-04 (2004). A. Gupta, L. Whitman, R.K. Agarwal, Supply chain agent decision aid system (SCADAS), Proceedings of the 2001 Winter Simulation Conference, 2001, pp. 553 559. D. Frey, T. Stockheim, P. Woelk, P. Zimmermann, Integrated multi-agent-based supply chain management, Proceedings of the Twelfth IEEE International Workshops on Enabling Technologies: Infrastructure for Collaborative Enterprises, 2003. H.S. Nwana, Software agents: an overview, Knowledge Engineering Review 11 (3) (1996) 205 244. N.R. Jennings, K. Sycara, M. Woodridge, A roadmap of agent research and development, Autonomous Agents and MultiAgent Systems 1 (1) (1999) 7 38. Z. Ren, C.J. Anumba, Multi-agent systems in construction state of the art and prospects, Automation in Construction 13 (2004) 421 434. O.B. Kwon, K.C. Lee, MACE: multi-agents coordination engine to resolve conflicts among functional units in an enterprise, Expert Systems with Applications 23 (2002) 9 21. N.M. Sadeh, D.W. Hildum, D. Kjenstad, A. Tseng, MASCOT: an agent-based architecture for dynamic supply chain creation and coordination in the Internet economy, Production Planning and Control 12 (3) (2001) 212 223. P. Lou, Z.D. Zhou, Y.P. Chen, A. Wu, Study on the multiagent-based agile supply chain management, International Journal of Advanced Manufacturing Technology 23 (2004) 197 203. T. Khedro, AgentCAD: a distribution cooperative CAD environment, Information Representation and Delivery in Civil and Structural Engineering Design, Civil-Comp Press, Edinburgh, 1996, pp. 15 19. C.J. Anumba, O.O. Ugwu, L. Newnham, A. Thorpe, Collaborative design of structures using intelligent agents, Automation in Construction 11 (2002) 89 103. F. Pena-Mora, C.Y. Wang, Computer-supported collaborative negotiation methodology, Journal of Computing in Civil Engineering 12 (2) (1998) 64 81.

430

X. Xue et al. / Automation in Construction 14 (2005) 413430 [40] R. Vrijhoef, L. Koskela, G. Howell, Understanding construction chains: an alternative interpretation, Proceedings of 9th Annual Conference International Group for Lean Construction, Singapore, 2001. [41] W.J. OBrien, K. London, R. Vrijhoef, Construction supply chain modeling: a research review and interdisciplinary research agenda, Proceedings of 10th Annual Conference International Group for Lean Construction, Brazil, 2002. [42] M. Muya, A.D.F. Price, A. Thrope, Contractors supplier management, Proceedings of a Joint CIB Triennial Symposium, Cape Town, vol. 2, 1999, pp. 632 640. [43] N. Fisher, R. Morledge. Supply chain management. Best Value in Construction. Blackwell Science, RICS Foundation (2002). [44] S. Mohamed, Web-based technology in support of construction supply chain networks, Work Study 52 (1) (2003) 13 19. [45] F. Andreas, Supply chain coordination by means of automated negotiations, Proceedings of the 37th Hawaii International Conference on System Sciences, 2004. [46] C.M. Jonker, J. Treur, An agent architecture for multi-attribute negotiation, in: B. Nebel (Ed.), Proceedings of the 17th International Joint Conference on AI, IJCAI01, Washington, USA, 210 August, 2001, pp. 1195 1201. [47] M. Barbuceanu, W.K. Lo, A multi-attribute utility theoretic negotiation architecture for electronic commerce, Proceedings of the Fourth International Conference on Autonomous Agents, 2000. [48] J.S.K. Lau, G.Q. Huang, K.L. Mak, Impact of information sharing on inventory replenishment in divergent supply chains, International Journal of Production Research 42 (5) (2004) 919 941. [49] M. Wooldridge, N.R. Jennings, Intelligent agents: Theory and practice, The Knowledge Engineering Review 10 (2) (1995) 115 152.

[28] J.U. Min, H. Bjornsson, Agent based supply chain management automation, Proceedings of the Eighth International Conference on Computing in Civil and Building Engineering (ICCCBE-VIII), 2000, pp. 1001 1006. [29] Z. Ren, C.J. Anumba, O.O. Ugwu, The development of a multi-agent system for construction claims negotiation, Advances in Engineering Software 34 (2003) 683 696. [30] K. Kim, B.C. Paulson, An agent-based compensatory negotiation methodology to facilitate distributed coordination of project schedual changes, Journal of Computing in Civil Engineering 17 (1) (2003) 10 18. [31] J. Rosenschein, G. Zlotkin, Rules of Encounter, MIT Press, Cambridge, MA, 1994. [32] J.M. Eilenberg, Dispute Resolution in Construction Management, University of New South Wales Press, Sydney, Austrilia, 2003. [33] R. Keeny, H. Raiffa, Decisions with Multiple Objectives: Preferences and Value Tradeoffs, John Wiley Sons, 1976. [34] J.K. Kim, S.H. Choi, A utility range-based interactive group support system for multiattribute decision-making, Computer and Operations Research 28 (2001) 485 503. [35] A.D. Youngblood, T.R. Collins, Addressing balanced scorecard trade-off issues between performance metrics using multiattribute utility theory, Engineering Management Journal 15 (1) (2003) 11 17. [36] M. Kelly, M.C. Thorne, An approach to multi-attribute utility analysis under parametric uncertainty, Annals of Nuclear Energy 28 (2001) 875 893. [37] H.S. Nwana, D.T. Ndumu, L.C. Lee, J.C. Collis. ZEUS: A Toolkit for Building Distributed Multi-Agent Systems. (2003). http://agent.aitiaai/download. [38] PLC BT, The ZEUS agent building toolkit methodology documentations. (1999). [39] F.T. Edum-Fotwe, A. Thorpe, R. McCaffer, Organizational relationships within the construction supply-chain, Proceedings of a Joint CIB Triennial Symposium, Cape Town, vol. 1, 1999, pp. 186 194.