Evolving software agent behaviours

As the scale and complexity of telecommunication networks have expanded to accommodate the increase in services offered to customers, a similar increase in the size of network management software has occurred. The software agent paradigm is a promising technique for developing complex software systems. By evolving control code for agents with a genetic algorithm (GA), it promises to allow agents to adapt their behaviour to local network conditions, and hence re-program themselves as circumstances change. Current work has focused on studying the performance of evolved agent behaviours on two search problems