How an agent makes decisions while keeping responsiveness

Faced with the need to solve problems of an inherently distributed nature, we have implemented a distributed computational system, which enables an intelligent integration of subsystems in the area of supervision and control of processes, that usually would be controlled by human operators. The main goal is the integration of different expert systems (possibly pre existing ones) enabling the cooperation among them, mainly in industrial environments. The transformation of expert systems into distributed agents of a cooperative community is achieved by the addition to each expert system of a new software layer-the cooperation layer-which is instantiated with information specific to those agents at the very right moment of the creation of the multi agent community. To promote the collaboration among the agents included in the community, the cooperation layer, which may be seen as a knowledge based system, makes it possible to control the activity of the underlying expert system, as well as the cooperation with the other agents, while still being responsible for the coordination of both these two activities. In order to achieve these three features, there is inside each cooperation layer, a blackboard based architecture enabling arriving messages to report other agents' current activity, to trigger knowledge sources which modify the goal intentions expressed in the agent's agenda, and hence, dynamically adopting all agent activities in order to pursue the system's overall goal.