An agent -based negotiation model for the sourcing of construction suppliers

Bidding is a critical decision process as it would determine whether a contractor can win a construction contract at a reasonable profit margin. Among various factors affecting the bidding decisions, the selection of suitable suppliers for the provision of various construction materials is one of the most important aspects in ensuring bid and project success. Because of the shortage of time at the bidding stage, the negotiation with suppliers is normally carried out in a crude and unstructured manner, and the terms and prices for material supply could be far from ideal. Although many researchers have proposed mathematical programming models to facilitate the sourcing decisions, these models are static and do not take the qualitative and quantitative factors into consideration very well. These models, therefore, cannot respond to the dynamic market conditions well enough.

Automated negotiation using the Multi-Agent System (MAS) concept can reduce the response time and maximize the expected profit should a suitable negotiation model be designed. The agent technology is suitable to the supplier sourcing, not only because it can facilitate a series of agents mimicking the participants involving in the negotiation process to achieve their goals in the real world, but also the MAS matches the fragmented nature of the construction industry.

This thesis describes the work of design of an agent-based negotiation model for the sourcing of construction suppliers. In this study, a literature review was first carried out to investigate the existing tools in supplier selection as well as selection criteria adopted by decision makers in construction industry. Then based on the design of a new parallel bargaining protocol, a conceptual negotiation model is designed while thoroughly studying the characteristics and problems existing in the sourcing of construction suppliers. The Zeuthen's concession strategy was adopted as the risk evaluation strategy in the negotiation model. Bayesian learning model has been integrated into the whole negotiation model to improve the efficiency. A prototype was built using JADE agent building toolkit based on Java.

Finally the model is assessed using the prototype and Face validation to evaluate the quality of the designed model. Results of the assessment show that this model can improve the efficiency of the negotiation to reach the win-win goal, as well as it can significantly improve the transparency. Furthermore, the results have showed that this model can also be utilized in the multi-attribute decision problem in a distributed environment such as supply chain scheduling, concurrent engineering and so on. This study has also proved the importance of a comprehensive study of negotiation model for the construction supplier sourcing.