Knowledge Compilation for Interactive Design of Sequence Control Programs
Yasuo Namioka, Toshikazu Tanaka · Industrial and Engineering Applications of Artificial Intelligence and Expert Systems · 2022
This paper presents a knowledge compilation method which improves the performance of planning-based interactive systems. The method automatically generates state space models from domain knowledge . It is generally difficult to adopt planning to aid interactive design because designers require quick response time in their work, and because design problems have more than several million states. We have been developing a planning-based visual Programming System for Sequence control program SpeciFiCation Design (VIPS/S-SFCD). The System addresses the difficulties by (i) dividing the state space models into the machine state space and the equipment state space, and by (ii) automatically generating the machine-level state space models from domain knowledge in short time without errors. The knowledge compilation also reduces search space in planning and the number of planning steps. Systems adopting our approach are capable of designing interactively, improving the reliability and quality of control specifications, and rapidly adapting with modification of a complex plant.