Real-time Autonomous Control

In this paper the concepts and models for intelligent control in real-time are given. The application of intelligent components in a (real-time) control scheme leads to a hierarchical system architecture, applying various AI-related techniques at different levels. The method to be applied depends on the required precizion and quantization and the knowledge available. Such an intelligent control scheme has to be embedded in a real-time environment, requiring specific features which should be apparant in a real-time expert system. Those specific requirements to apply AI related techniques in real-time led to the development of a testbed environment: DICE. This environment offers a blackboard solution using several knowledge sources attached to this blackboard. The knowledge sources operate in real time, using blackboard communication and progressive reasoning. Rule compilation is used for fast inferencing. The knowledge description includes the use of multivalued logic at the rule level. Two examples of real-time knowledge-based control using fuzzy logic are given