Formal Stepwise Development of Scalable and Reliable Multiagent Systems

Abstract

This chapter considers the coordination aspect of large-scale dynamically-reconfigurable multi-agent systems in which agents cooperate to achieve a common goal. The agents reside on distributed nodes and collectively represent a distributed system capable of executing tasks that cannot be effectively executed by an individual node. The two key requirements to be met when designing such a system are scalability and reliability. Scalability ensures that a large number of agents can participate in computation without overwhelming the system management facilities and thus allows agents to join and leave the system without affecting its performance. Meeting the reliability requirement guarantees that the system has enough redundancy to transparently tolerate a number of node crashes and agent failures, and is therefore free from single points of failures. The Event B formal method is used to validate the design formally and to ensure system scalability and reliability.