Optimal Task Allocation for Human-Machine Collaborative Manufacturing Systems

Abstract

Human-machine collaborative manufacturing systems consist of human operators and automated machines. They cooperate with each other to accomplish complex tasks that are difficult for either the human or machine alone. This letter considers an optimal task-allocation problem for human-machine collaborative manufacturing systems, where various tasks are dispatched to human operators and automated machines to achieve optimal joint human-system performance. Designing such optimal task allocation is challenging because of the stochastic hybrid feature of manufacturing processes, as well as varying human performance caused by physical fatigue. To address this challenge, we first model human fatigue as a continuous-time Markov decision process, which is capable of capturing stochastic uncertainties on fatigue dynamics under different task assignments. A novel controlled stochastic petri net is then proposed to model the manufacturing process, in which both time- and event-driven dynamics can be regulated by task allocation between the human and machine. Under mild assumptions, we show that the optimal task-allocation problem under the proposed human manufacturing framework can be solved by linear programming. Simulation results of a four-part assembly process are used to verify our theoretical findings.