A Scalable and Efficient Backpressure-Based Scheduling Framework for Supply Chain Networks

Supply chain networks have proven to be rather important and delicate to seemingly small perturbations of their operations. Although scheduling has been extensively studied in logistics systems, there are several remaining open challenges regarding scalability, stability, and other quality indices. In this work, we present a novel framework denoted as BackPressure-style Packet transfer algorithm for Logistics Systems (BPLS) for making jointly optimal routing and scheduling decisions in freight networks. The proposed approach is based on the backpressure algorithm and maximum weight matching, which have been extensively applied for optimal routing and scheduling of data packet transmissions in communications networks. Our goal is to develop a broader optimal transfer process for freight networks consisting of multiple entities, such as last-mile companies, freight subcontractors, etc., addressing all the previously mentioned challenges and in addition, allow for setting different optimization goals regarding the offered quality of service. Special features of freight networks such as the limited capacities of both storage places and transportation means along with the time-varying availability of the latter are considered by means of integrating pressure functions in the original backpressure approach. We provide extensive simulations with evaluation and comparison results on the performance of the approach, demonstrating its potential to improve up to more than 100\times on the traditional BP algorithm. In addition, we have incorporated an implementation in an operational information system to assess the potentials of BPLS with multiple interested stakeholders. Through simulations and the actual evaluation, we were able to show how the framework can be used to provide long and short term decisions for optimizing holistically freight networks.