A human-robot collaborative order picking system with ergonomic considerations: A novel mathematical model and machine learning approach

摘要

• Develop a warehouse order-picking model with ergonomic and cost considerations. • The model promotes human-cobot collaboration in warehousing operations. • Develop a new Machine Learning approach to solve complex problems. • Identify and evaluate factors affecting system performance. • Propose integrating warehouse and order-picking system design. Order Picking Systems (OPS) play a critical role in warehouse operations, particularly in markets where delivery speed is a key competitive factor. While traditional OPSs are labor-intensive, integrating collaborative robots (cobots), such as automated mobile robots (AMRs) and robotic manipulators, offers potential improvements in efficiency and ergonomics. This study proposes a mathematical model for a Human-Robot Collaborative OPS (HRC-OPS) that optimally determines the number and type of cobots to deploy. The objective is to minimize operational costs while maintaining ergonomic safety by limiting the REBA (Rapid Entire Body Assessment) index. In addition to robot allocation, the model optimizes item placement within the warehouse to enhance system performance and worker well-being. As the model aims to reflect real-world warehouse operations, it becomes too complex to solve using exact methods. However, since the model seeks to guide managers’ strategic choices during the design phase, an exact solution is not a priority. Therefore, a machine learning (ML) approach was developed to extract patterns from high-quality solutions and provide actionable managerial insights. Among six tested ML algorithms, XGBoost showed the highest accuracy in identifying effective configurations. Results from a case study demonstrate that cobots can significantly enhance OPS performance. However, the effectiveness of specific robot types depends on system characteristics, such as demand frequency and physical attributes. Moreover, strategic item placement has a direct influence on both performance and ergonomic outcomes, particularly for frequently ordered items or those that are unsuitable for robotic picking, offering practical guidance for warehouse managers.