A learning approach to enhancing machine reconfiguration decision-making games in a heterarchical manufacturing environment

Business process management (BPM) emerges as a promising guiding principle and technology for integrating existing assets and future deployments. BPM provides mechanisms to transform the behaviours of disparate and heterogeneous systems into standard and interoperable business processes, aimed at effectively facilitating the conduct of factory system integration. On the shop floor, BPM requires machine-level controllers to be architected in a heterarchical fashion. Past work by the authors introduced a heuristic non-cooperative game theoretic planning technique for an autonomous machine controller to frame a decision about an impending reconfiguration (i.e. setup change) in a heterarchical manufacturing environment. The work described in this paper extends the authors’ previous work by employing a reinforcement learning approach for specifying the payoffs in reconfiguration games through capturing the effects of a sequence of reconfiguration decisions. Consequently, the controller can autonomously learn the long-term implications of decisions, and ultimately improve its decision-making process in manufacturing.