Enabling Autonomous Decision-Making in Manufacturing Systems through Preference Fusion

Decision analysis (DA), a well-established discipline in business and engineering, is entering another domain of application due to the advent of Industry 4.0. DA enables optimal decisions by finding system parameters that maximize the utility, or in the presence of uncertainty the expected utility, from the attributes of a system. Whether there is a single decision maker or all decision makers have uniform preferences, determining risk behavior and the resulting utility is well developed in the existing literature. However, variability in preferences has not been satisfactorily addressed. This gap gains added significance in the face of the demands of Industry 4.0 where cyberphysical production systems must drive autonomous decision-making on the factory floor. The decisions must accommodate a distribution of customer and designer preferences, including production auditors within the organization. This article provides a novel framework and develops a closed-form approximation for expected utility in the presence of uncertainty in both attributes and preference behaviors. The value of this approach is demonstrated in the assembly of parts in a cyberphysical production system of an automotive manufacturing plant. The comparison of corrective assembly using the proposed method with existing random assembly approaches shows significant improvements.