Online Heuristics for the Dynamic Assignment of Workers in Contextaware Manufacturing Systems with Random Times and Fluctuating Demands

Worker assignment problems are encountered in many production systems and are known to greatly influence system performances. We focus on dynamic problems where the system is characterized by unexpected and fluctuated demands and random times. Workers are multiskilled and must be assigned to machines according to the need. In order to increase reactivity, the assignment decisions in context-aware manufacturing systems can be made in real-time, according to the evolution of the system state. In this article, we suggest two heuristic approaches that determine online, depending on the system state, the next machine to which a worker should perform an operation each time he/she becomes idle to reduce the mean flow time. The first heuristic (H1) is based on a multi-criteria decision making using TOPSIS (Technique for Order Performance by Similarity to Ideal Solution) while the second one (H2) is based on the identification of the most important machines. These approaches are generic enough to be used to solve a wide variety of workers assignment problems. The weight values associated with each criterion used in H1 and a set of numerical thresholds used in H2 are defined offline, using simulation optimization, according to the system’s operating conditions and to the desired objective. The approaches are assessed on a job shop system with multiskilled workers and their performance are compared to efficient assignment rules selected from the literature. After optimization, these heuristics are found to yield significantly better results