A multi-agent system is a group of software or hardware agents that cooperate or compete to achieve individual or shared goals. A method for developing a multi-agent system must be capable of modelling the concepts that are central to multi-agent systems. These concepts are identified in a review of Agent Oriented Software Engineering methodologies. The rigorous development of complex systems using formal methods can reduce the number of design faults. Event-B is a formal method for modelling and reasoning about reactive and distributed systems. There is currently no method that guides the developer specifically in the modelling of agent-based concepts in Event-B. The use of formal methods is seen by some developers as inaccessible. This thesis presents an Incremental Development Process for the development of multi-agent systems in Event-B. Development following the Incremental Development Process begins with the construction of informal models, based on agent concepts. The informal models relate system goals using a set of relationships. The developer is provided with guidance to construct formal Event-B models based on the informal design. The concepts that are central to multi-agent systems are captured in the Event-B models through the translation from the goal models. The Event-B models are refined and decomposed into specifications of roles that will be performed by the agents of the system. Two case studies illustrate how the Incremental Development Process can be applied to multi-agent systems. An additional aid to the developer presented in this thesis is a set of modelling patterns that provide fault-tolerance for Event-B models of interacting agents.
展开▼
