文摘
Workflow adaptation is important for enterprises that are adjusting their functioning to deal with change; however, frequent adjustments to workflow are a costly and time-consuming process, which hinders the evolution of enterprise activities. Thus, there is a major need to enhance workflow flexibility and to automatize its adaptation to cope with these issues. In this paper, we introduce a novel approach for a self-adaptive workflow system. We propose an algorithm, which combines reinforcement learning and forecasting to enhance automatic workflow adaptation. The proposed solution integrates agents using Q-learning in a distributed way so as to determine the appropriate policy to adapt the workflow. It further applies auto-regressive integrated moving averages to predict changes in the workflow environment. For this purpose, we developed a collaborative multi-agent system to control these adaptations. Our major contribution consists of adapting the workflow in distributed way, whereby each agent adapts a workflow part while collaborating with other agents. A household appliances assembly plant is simulated to validate the proposed solution. The obtained results illustrate the applicability of the proposed solution for enhancing workflow adaptation and improving responsiveness to change. The effectiveness of the proposed approach is evaluated by means of a comparison to a baseline approach.