Abstract:
|
This article shows how a conventional problem solver is extended to have a set of agents, each of them based on the former problem solver, that cooperate in a social framework. The evolution and differences from the single resolution system towards the distributed one are explained. A generic tool for the design and implementation of systems using this model has been developed. A remarkable feature of the resulting system is the independency of the knowledge of the problem solving, from the architecture that supports the society. Problem resolution is goal driven. Goals are solved by agents with their own methods, or by submitting them to other agents. During a method execution new subgoals can be generated, starting new problem resolution cycles. The architecture that supports the model is presented too. A working example is described, showing the tool use. |