Automatic Design of Dispatching Rules for Real-time optimization of Complex Production Systems *

Dispatching rules have proven to be very useful in practice for real-time optimization of production systems. Due to the complex and dynamic nature of manufacturing environment, there are no universal rules that are superior in all situations. The task of manually designing dispatching rules is not trivial and time consuming. Recently, genetic programming becomes a popular method to facilitate the design task by automatically generating dispatching rules. This paper develops an efficient hybrid method, which achieves an automatic design of dispatching rules for real-time optimization of complex production systems. We first formulate the evaluation problem as a ranking and selection problem, and then propose a new evaluation method for genetic programming by simulation optimization. The proposed method can be applied for evaluation of candidate heuristics to enhance the search efficiency of genetic programming. Numerical results show the efficiency and effectiveness of the proposed method.