Chapter 1
Introduction
C. J. Anumba, O. O. Ugwu and Z. Ren
1.1 Background
Collaborative working in construction remains a major precursor to
achieving improved productivity resulting in direct business benefits to construction
organisations. It is therefore imperative that researchers continue
to explore state-of-the-art paradigms, techniques and technologies that
would enhance efficient delivery of constructed products as measured using
various metrics such as: quality, cost (life-cycle cost), safety and sustainability.
The challenge is how to continuously improve the activities associated
with product delivery including design, procurement (i.e. tender
appraisal and construction material purchasing), specification, construction,
project management and knowledge management across construction
organisations.
Multi-agent systems (MAS) are a fast developing information technology
(IT), where a number of intelligent agents (IA), representing real world entities,
co-operate or compete to reach the desired objectives of their owners.
The increasing interest in MAS is because of its ability to provide robustness
and efficiency, to allow inter-operation of existing legacy systems and
to solve problems in which data, expertise or control are distributed.
The general goal of MAS is to create systems that interconnect separately
developed agents, thus enabling the ensemble to function beyond the capabilities
of any singular agent in the set-up (Nwana and Ndumu, 1999). This
is because the limited knowledge, computing resources and perspectives
limit the capability of a singular agent. Thus, if a problem domain is
particularly complex, large or dynamic (such as in the construction industry),
then the only way it can be reasonably addressed is to develop a number
of functionally specific and modular components (agents) that specialise
in solving a particular problem. This decomposition allows each agent to
use its best knowledge for solving the particular problem. Thus, when
interdependent problems arise, the agents need to co-ordinate or collaborate
with one another to ensure that interdependencies are properly
managed from different perspectives.
Fragmentation is one of the major problems in the construction industry.
Different project participants, often geographically distributed, need to
co-operate and collaborate to perform various construction activities. This
makes the need for effective information and communication technologies
acute. Although various ITs such as expert systems, databases and the
Internet have greatly enhanced the productivity of the industry, none of
them can solve the fragmentation problem of the industry. The additional
problems posed by the use of heterogeneous software tools are well known
and need to be overcome by the adoption of new approaches such as the
use of IA. IA consist of self-contained knowledge-based (k-b) systems that
are able to tackle specialist problems and which can interact with one
another (and/or with humans) within a collaborative framework (Anumba
et al., 2002).
Construction problems also involve open and dynamic problems where
the structure of the system itself is capable of changing dynamically. The
characteristics of such a system are that its components are not known in
advance, can change over time and can consist of highly heterogeneous
agents implemented by different people, at different times, with different
software tools and techniques. These capabilities require that agents be able
to inter-operate and co-ordinate with each other, and to learn from one
another and the environment. As a result, MAS have considerable potential
to address some of the fragmentation problems of the industry.
1.2 Structure of the book
This book consists of chapters that describe current developments and
future directions in the theory and application of IA and MAS in the
Architecture, Engineering and Construction (AEC) sector. The book reflects
an effort from an international network of construction IT researchers (and
related disciplines such as computing), investigating different aspects of
agent theory and applications.
The chapters contributed cover different perspectives and application areas.
The reported research projects represent significant efforts to harness emerging
technologies such as IA and MAS for improved business processes in the
AEC sector. Although the chapters in the book can be read in any order, they
can be broadly grouped into two areas: (a) agent theoretical foundations in
Chapters 2ā4; and (b) construction applications ā Chapters 5ā11. However,
the authors also discuss relevant theoretical underpinnings in the respective
application areas. Chapter 12 summarises the book and draws a number of
conclusions.
Chapter 2 focuses on some perspectives and overview of IA in civil
engineering. It explores the evolutionary dimensions of systems in engineering
design and describes the metamorphosis from first-generation (k-b) systems
to current systems that are underpinned by evolutionary computation. It also takes a quantum leap into the future evolution of agent-based
systems and discusses potential applications of evolutionary computing
techniques to improve agent behaviours in problem solving. The chapter
differentiates between IA and the emerging sapient agents, and identifies
additional attributes for agent classifications. Using these attributes, it
proposes a definition of an IA and a taxonomy based on: interaction range,
interaction depth, learning and knowledge, structure and quantity. This
contributes new dimensions to the evolving classification of agents within
the research community.
Chapter 3 discusses theoretical foundations and other fundamental
aspects of agent research and development. These include agent taxonomy,
co-ordination, negotiation and learning, amongst others. Negotiation
theories are discussed in detail in Chapter 4. This recognises the fact that
negotiation as well as other aspects such as co-ordination and interaction
are pivotal in a MAS environment. The two predominant models and paradigms
in mainstream negotiation studies ā economic and game-theoretic
models are described with a clear articulation of the most appropriate
situation for their use.
Chapter 5 presents the first MAS application, which addresses agentbased
collaborative design. It includes methodological issues in MAS development,
computational modelling of the identified business processes,
design knowledge modelling and representation issues that underpin fully
automated peer-to-peer negotiation as well as other reactive and deliberative
behaviours and characteristics of agents in a MAS design space. The
collaborative design of a steel-framed light industrial building is used to
illustrate the potential of the resulting prototype system.
The application of agent technologies to construction claims negotiation
is the subject of Chapter 6. The chapter presents all the main aspects of this
domain including modelling the practical claim negotiation problem,
analysing negotiatorsā roles in an agent-based system, designing the negotiation
protocols and developing agent negotiation strategies. A particular
agent-learning approach is also integrated into the negotiation mechanism.
The chapter also provides a general methodology for the development of
MAS negotiation mechanism in construction. This is particularly important
because very little has been published on this aspect. A prototype system,
MASCOT, is presented which encapsulates the methodology and illustrates
the potential of agent-based construction claims negotiation. Examples of
the working of the system are also included.
Chapter 7 describes the use of agents to specify and procure construction
products. The difficulties associated with aggregating product information
from heterogeneous sources are highlighted and the approach adopted to
overcome these described. A prototype system, APRON, which facilitates
product specification and procurement is presented and its functionality
illustrated using an example.
Chapter 8 presents the application of agents in bidding, during which the
agents facilitate procurement solutions through the sourcing and purchase of
building materials, equipment, and supplies in a virtual marketplace and in
compliance with the German standards and regulations. The resulting prototype
application demonstrates extensions of the functionality provided by
agent development platforms to address construction-specific issues, such as
the use of Public Key Infrastructure (PKI) to offer security to agents participating
in the tendering process. The chapter also discusses various software
models for virtual AEC bidding.
Information search and retrieval is the focus of Chapter 9, which
discusses the application of agents to provide a key generic support service
that cuts across various application domains. Information search, retrieval
and delivery to users using IA provide a solution to the problems of information
overload as more context-specific and user-oriented information
can be delivered. This chapter describes how this can be done.
Chapter 10 explores the application of agents to standards processing. It
demonstrates agent application to standards evaluation and validation,
which cut across various domains in engineering design. It starts with
a review of related work on standards processing and presents a distributed
framework for agent-based standards processing. The proposed MAS tackles
scalability and other problems, with each agent using a different representation
and reasoning method rather than a single, very general method.
Chapter 11 discusses another application of agent-based systems in
procuring construction materials. It gives a detailed analysis of construction
supply chain operations including several existing problems. The chapter
then identifies how MAS can address these perennial problems and contribute
to streamlining the business processes in materials procurement and
delivery. It also demonstrates another extension of the contract-net protocol
(Smith, 1980), in the context of agent application for construction materials
procurement. The chapter focuses on materials procurement as part of
the product delivery process at micro-level operations.
In deploying new technology in any sector, it is important to address the
technical implementation issues (technology-related factors) as well as
non-technical issues (soft or human factors). This view of organisations as
socio-technical systems has been the subject of considerable research in
mainstream computing (specifically software) engineering. It is also an area
that demands focused research by researchers in the AEC sector, because
both technical and social factors affect the success or failure of an IT project
(as measured by implementation, adoption and depth of usage to enhance
productivity through process innovation). In the light of this, Chapter 12
discusses some of the issues raised in various chapters of the book. It highlights
the socio-technical dimensions and draws some conclusions on agent
applications in construction.
We hope that after reading this book, readers would have a deeper
understanding of theoretical and practical dimensions that underpin MAS
(and other aspects of intelligent systems) research in construction. We also
hope that this would generate new interest in the MAS paradigm, as well as
other state-of-the-art IT and computing techniques.
References
Anumba, C. J., Ugwu, O. O., Newnham, L. and Thorpe, A. (2002) āCollaborative
design of portal frame structures using intelligent agentsā, Automation in
Construction, 11(1): 89ā103.
Nwana, H. and Ndumu, D. (1999) āA perspective on software agents researchā, The Knowledge Engineering Review, 14(2): 125ā42.
Smith, R. G. (1980) āThe contract net protocol: high level communication and
control in a distributed problem solverā, IEEE Transactions on Computers,
C-29(12): 1104ā13.
Chapter 2
Intelligent agents
Fundamentals
T. Arciszewski, Z. Skolicki and K. De Jong
2.1 Introduction
The progress of our civilisation is mostly driven by the changes in science. First, in the era of agrarian societies, the progress was driven by the advances in agriculture. Next, in the industrial societies, the progress was a result of improvements in manufacturing. Today, we are living in the post-industrial societies and their evolution results mostly from the advances in information and knowledge processing. For this reason, our times are often called the era of Information Technology Revolution, as a way of describing the complex synergistic process of changes occurring in the society and within engineering. These changes are driven by new sciences, technologies and tools, all related to information and knowledge processing.
In Civil Engineering, the Information Technology Revolution is reflected in the emergence of knowledge-based (k-b) systems during the last 20 years. Such systems are intended for various decision-making purposes including design, planning, maintenance of infrastructure systems, etc. Recently, there is a growing interest in a new category of such systems called āIntelligent Agentsā (IAs). IAs constitute the most advanced form of (k-b) systems and many researchers and practitioners expect them to become the dominant form of (k-b) systems, which will soon change civil engineering practice entirely. At the moment, the notion of IAs is still not fully understood, both within Computer Science and Civil Engineering communities. Therefore, various definitions of IAs are introduced and discussed in Section 2.3. Our definition is thoroughly explained and justified in the same section. However, it is given here in order to provide sufficient context for the introduction. It is as follows:
An intelligent agent is an autonomous system situated within an environment. It senses its environment, maintains some knowledge and learns upon obtaining new data and, finally, it acts in pursuit of its own agenda to achieve its goals, possibly influencing the environment.
(Skolicki and Arciszewski, 2003a)
The process of changes in Civil Engineering within the domain of (k-b) systems can be understood as a process of evolution. As such, it can be considered in the context of Directed Evolution (Clarke, 2000). This term reflects an emerging Engineering Science that deals with the analysis of the evolution of engineering systems over periods of time. The basic premise of Directed Evolution is that the evolution of engineering systems is driven by objective patterns of evolution. Eight such patterns have been identified as a result of studies of the evolution of many engineering systems developed in various countries over long time periods. In the case of (k-b) systems, four patterns seem to be particularly relevant. Therefore, they will be used to explain why IAs are emerging and why this process is so significant. These relevant patterns of evolution and implications on the evolution of (k-b) systems are discussed later.
2.1.1 Evolution towards decreased human involvement
When the evolution of an engineering system is considered over a period of time, it can be easily observed that the required human involvement is systematically decreasing. For example, when a car is considered, every year improvements are made to reduce human involvement in the operations of the carās engine, brakes or steering. Similarly, each new computer system design attempts to simplify and reduce human involvement.
In the case of (k-b) systems, this evolution pattern means that with time their use will require less and less human input. The last 20 years of evolution of (k-b) systems have already confirmed the validity of this evolution pattern. When the first (k-b) systems were developed they were written in symbolic languages (LISP or Prolog), their user interfaces were poor or non-existent, and their use required tremendous effort. Gradually, various shells for building (k-b) systems with excellent user interfaces have been developed, and their introduction significantly reduced the amount of effort required to use such systems. Obviously, the natural trend is in the direction of autonomous systems. That may mean the reduction of human involvement in the case of the majority of (k-b) systems and the emergence of highly autonomous systems. This second group may become gradually dominant. The systems in this group would require only very limited human involvement to perform their functions and that involvement most likely will be on the level of meta-rules representing abstract knowledge. Ultimately, the pattern of Evolution towards decreased human involvement means in this case that IAs are the natural goal of evolution of (k-b) systems and that their emergence should be simply expected. The gradual shift from systems totally controlled by humans to systems interacting with humans and ultimately to autonomous systems is illustrated in an abstract way in Figure 2.1.
Figure 2.1 Evolution towards decreased human involvement.
Figure 2.2 Evolution towards the micro-level and increased use of fields.
2.1.2 Evolution towards the micro-level and increased use of fields
This pattern of evolution can be explained discussing the evolution of computing devices. In this case, the ...