The 'two-hop principle' : stability of MAS in communications networks

A joint university / industry collaboration is investigating the deployment of MAS technology to manage wireless communications networks. These systems are very large and are situated in a chaotic environment which suggests the use of distributed intelligent control that aims to continually improve, whilst not expecting to optimise, network performance. A key problem is ensuring the stability of these large, distributed control systems. A sequence of management systems have been developed over four years from which we have abstracted our 'two-hop principle'. We conjecture that the principle has wider application to multiagent network management systems. The generic principle is described and experiments are reported to illustrate the stability observed.