Negotiation within a multi-agent system for the collaborative design of light industrial buildings C.J. Anumba a, * , Z. Ren a , A. Thorpe a , O.O. Ugwu b , L. Newnham c a Department of Civil and Building Engineering, Loughborough University, Loughborough LE11 3TU, UK b Department of Civil Engineering, The University of Hong Kong, Hong Kong, People’s Republic of China c Touch Clarity Ltd, 28, Bruton Street, London W1J 6QW, UK Received 6 December 2001; accepted 6 March 2003 Abstract This paper is a review of intelligent agents with respect to their use within the Agent-Based Support for The Collaborative Design of Light Industrial Buildings (ADLIB) project. In the ADLIB project, the core objective is to develop a multi-agent system (MAS) framework for the representation of activities and processes involved in collaborative design of light industrial buildings. This includes the planning and fabrication of steel structural components. ADLIB intelligent agents are concerned with modelling action and knowledge in a collaborative environment. The design process that ADLIB’s agents are trying to automate is the interaction and negotiation between specialist design team members. Each team member with a different area of expertise will be primarily concerned with his own area of interest. This paper starts with an introduction to intelligent agents. It then moves on to a discussion of agent classification systems and negotiation theories and their applications in MAS. The last section analyses the needs of agents within the ADLIB project. A negotiation protocol and strategy are then presented. q 2003 Elsevier Science Ltd. All rights reserved. Keywords: Collaborative design; Multi-agent system; Negotiation mechanism 1. Introduction There is growing interest in the adoption of colla- borative working practices in the construction industry as a means of addressing many of the problems caused by the fragmentation of the industry. A key aspect of collaborative working between the multi-disciplinary teams involved in construction projects is facilitating the flow of information across the heterogeneous software tools in use. Distributed artificial intelligence offers an innovative approach to overcome this problem. The ADLIB (Agent-based support for the collaborative design of light industrial buildings) research project [2–5,45] aims to investigate the use of intelligent agents to facilitate collaborative design and will focus initially on the design of light industrial buildings, particularly portal frames. This approach, which overcomes the problem of geographically distributed teams, is novel in the construction industry. It also goes beyond existing systems developed for construction in seeking to develop intelligent agents that are able to negotiate with one another to arrive at a mutually agreed design solution. Depending on the scale and complexity of a project, the project team may comprise a client, an architect, a structural engineer, a building service engineer and other members such as health and safety regulators and other specialist subcontractors [1]. The activities during the design stages involve a lot of negotiation and information/data exchange between these design groups. The core objective of the ADLIB project is to develop a multi-agent system (MAS) framework for the representation of activities and processes involved in the collaborative design of industrial buildings. Therefore, the ADLIB project will investigate the interaction protocols for intelligent agent negotiation in such a system. This paper first introduces the concept of intelligent agents, agent classification systems, agent organisation structures, and agent working domains. It then examines some of the major negotiation theories and their possible 0965-9978/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0965-9978(03)00038-3 Advances in Engineering Software 34 (2003) 389–401 www.elsevier.com/locate/advengsoft * Corresponding author. Tel.: þ44-1509-222-615; fax: þ 44-1509-223- 981. E-mail address: c.j.anumba@lboro.ac.uk (C.J. Anumba).