Digital Twins for Supply Chains


What is a Supply Chain Digital Twin? 



Supply Chains emerge as complex adaptive networks of interdependent companies. While companies know whom their immediate neighbours (buyers and suppliers) are, they often lack full supply chain visibility beyond these immediate connections. Yet somehow they must coordinate to deliver goods and services on time, whilst also reducing overall costs. This information asymmetry leads to well-documented emergent effects such as demand amplification, and systemic disruption risks. Information asymmetry results in wasted material, as each individual company in the network inflates inventory to act as a buffer against possible disruptions, with no access to the actual data. While methods such as Collaborative Planning, Forecasting and Replenishment have been proposed, they require companies to integrate individual ERP systems, or set up data patches, which require investment, time, and discipline. Cost becomes a barrier to the exchange of information, and investment becomes a risk as it perpetually ties companies together, preventing the formation of new alliances. A more fluid, flexible, and decentralised solution is needed to overcome information asymmetries. 


Digital Twins offer us a natural way to understand and model these systems, predict how they will behave and allow us to experiment with what-if scenarios, which can then be used to identify effective intevention mechanisms and policies to "nudge" the system to a globally applicable goal. So, a supply chain digital twin should encapsulate the majority of the data science lifecycle, from descriptive to prescriotive analytics. 


A Supply Chain Digital Twin could have multiple manifestations: It could be a virtual representation of a physical product moving across the value chain (closer to what a "traditional" digital twin would be), a process centric twin would include information flow such as transactions over the entire value chain, an "ecosystem digital twin" which models the wider environment and the extended complex system a company operates within. 


We have multiple projects in Supply Chain Digital Twins.  


Autonomous Digital Twins for Collaborative Logistics: This project explores how agents representing carriers can convince each other to share trucks. We create multiple digital twins that "chat" with each other to alert sharing opportunities, so they can reduce carbon on UK roads by using tucks more efficiently. See our latest demonstrator in AI UK 2023.  


ChattyTwins: Sponsored by the Alan Turing Institute, in this project we investigate how multiple autonomous agents representing individual companies in a supply chain can achieve better outcomes by collaborating. To achieve this we research three topics. First, how can we find relevant information in multi-agent systems with data privacy. Second, how can we remove outdated information in multi-agent systems. Third, how can we incentivise other agents to reach a better outcome together.


A Surveillance Digital Twin: Could we couple a supply chain digital twin, with a software agent that collects information about the wider system a company operates in? In this project we are seeking ways to incorporate wider systemic uncertainty into supply chain operations. 



Funded by: EPSRC, Research England, Alan Turing, Accenture

Collaborators: University of Oxford

Researchers: Liming Xu, Stefan Schoepf, Stephen Mak



Share This