Knowledge-based two-level control of a group of mobile executors

Abstract

Two interconnected sub-problems, i.e. scheduling of independent, non-preemptive tasks on unrelated executors as well as motion control of a group of moving executors performing the tasks, which form a two-level manufacturing operation system are investigated. As the performance index of the two-level system the makespan is assumed. For the motion control sub-problem a knowledge based pattern recognition procedure is used. The procedure is treated as a co-ordinator of movements of separate executors and allows avoiding their possible collisions. The knowledge-based pattern recognition problem is solved using the logic-algebraic method. An expert is assumed to be the source of knowledge about collisions, which is given in the form of logic expressions. Two heuristic solution algorithms for the two-level system are presented. The first algorithm ensures the current modification of solutions for the scheduling sub-problem during the control procedure of the two-level system. In the second one the on-line procedure is used, which enables determination of the best solution in the current decision step of the control procedure. Both algorithms are compared via computer simulation and examples of results are presented. A numerical example for both solution algorithms is also given. It concerns the selected process in a discrete manufacturing system.