A software agent distributed system for dynamic load balancing

Abstract

As distributed computer systems become more and more available, its role in science and society necessarily increases, and it will be necessary to consider ways in which these tools can be made to interact effectively. The development of what may be called of social engineering for communities of machines; i.e., the means to set up laws that will force those communities to behave in a certain beneficial way, to have machines controlling and sharing resources under a broadly acceptable doctrine of general goodness. Of interest is the orthogonal problem of formally describing the conditions for an efficient distribution of tasks to processors. A formal semantics is important because a rigorous understanding of how the communication in the system is processed is central to our goals. On the other hand, it will be quite useful to embed a version of the semantics in the machines themselves, so they can use it in their reasoning about their own and others actions. Deriving the rules of public behaviour or the rules of the game by which the machines adapt to a changing world, a logical theory to drive machine interaction emerges. A logical theory or logic program built on the interaction results of a set of software agents, making the MASES (Multi-Agent Sensor Effector System). The MASES design turns out to be highly sensitive, depending on the domain in which the agent interaction was taking place. To capture this sensitiveness and model the system one had to consider and borough concepts from fields such as genetic algorithms, classifier systems, and logic programming.