This paper presents a new data-driven model for parallel and efficient execution of logic programs. Applied to knowledge-based control of autonomous systems, it takes advatage of the fact that these systems use their set of rules in a special way.
A two-step method for constructing dataflow graphs is outlined. Its core part is the analysis of static data-dependencies, as they exist in the sets of rules of knowledge-based control.
The model does not restrict evaluation parallism in any way, i.e., the same amount of parallelism is exploited as by the (restricted) AND/OR-evaluation models of logic programming. Furthermore, control efficiency is increased by the model's purely data-driven evaluation scheme that saves any distribution of demands-for-evaluation.
* A short version of this paper was published with the same title in the Conference Proceedings of AI, Simulation and Planning in High Autonomy Systems, San Diego, 1996. pp. 323-330.
** Another short version of this paper was published with the title "Knowledge-based Dataflow Computation for Autonomous System" in the Proceedings of the 6th International Symposium on Measurement and Control in Robotics, Brussels, 1996, pp. 531-536.
Both papers extended by the Prolog-example of a control system for an autonomous vehicle, including examples for the calculated output and the complete structure and dataflow graphs for this example.
Ulrike Peiker, Martin Griebl