Automation solutions for E-commerce multi-item packing

Abstract

As Amazon continues to experience a rapid growth in its e-commerce business, fulfillment efficiency needs to through safe implementation of advanced technology to create a better customer experience. Amazon has heavily invested in automating its outbound product sortation process that merges picked items but has yet to develop automation for multi-item packing. Individual item manipulation has been proven very challenging to automate due to the over 500 million unique products offered. This thesis proposes a container manipulation solution that integrates industrial robotics and other equipment with upstream sortation technology to automate the packing process. A physical prototype was built to test the concept and measure proficiency in critical quality metrics such as item accuracy, product damage, and packing density/orientation. Additionally, an operational simulation for the system was developed to determine the optimal capacity sizing for the integrated sortation and packing system. Lastly, sensitivity analysis on a financial model was performed to optimize for the net present value (NPV) and payback period. After a series of controlled experiments and process improvements, the prototype produced promising results, given the rudimentary nature of the prototype. The system generated item accuracy defects at 2%, product damage defects at 2% and packing orientation defects at 17%. While these results are not adequate to be used in live operation, a development path to acceptable performance appears attainable. Furthermore, implementation of the technology would generate approximately and $100M in NPV across the global fulfillment network.