Data-driven scheduling optimization under uncertainty using Renyi entropy and skewness criterion

• Published in: Computers & industrial engineering 2018-12, Vol.126, p.410-420
• Main Authors: Wang, Zhiguo, Pang, Chee Khiang, Ng, Tsan Sheng
• Format: Article
• Language: English
• Subjects: Flexible manufacturing systems; Petri net; Renyi entropy; Resource cost uncertainty; Scheduling optimization; Skewness
• ISSN: 0360-8352; 1879-0550
• DOI: 10.1016/j.cie.2018.09.037

•A scheduling optimization problem under resource cost uncertainties is studied.•An integrated Renyi mean-entropy-skewness information criterion is proposed.•Data-driven approximations of the information criterion are developed.•Real industrial applications with diverse system scales and data types are studied.•Reduced average dispersion rates are achieved in all the computational experiments. In order to deal with the resource cost uncertainties, this paper introduces a Renyi mean-entropy-skewness (RMES) information criterion for the scheduling optimization problems in flexible manufacturing systems (FMSs). Motivated by potential limitations in the existing measures, this third-order information criterion is carefully integrated to be more general and more robust in representing the schedule dispersion under uncertainties. The RMES information criterion is estimated using data-driven techniques so that it does not rely on the assumptions of exact probability distributions which are usually unknown in practice. Modeled with Petri net (PN) and system state reachable graph (RG), the RG-based dynamic programming (DP) algorithm and an approximate dynamic programming (ADP) algorithm are presented to solve the proposed model. The effectiveness of the introduced information criterion is verified by both technical proofs and extensive simulation studies of systems with a wide range of scales and data types. A real stamping industrial case study is also conducted as a justification of the model’s practical applicability.