Adaptive AI Decision Agents for Holistic Supply Chain Optimisation: Merging Societal and Business Objectives

Please note the deadline for this opportunity has now passed and we will not be accepting further applications. 

Supply chain efficiency is a global problem that affects almost every person on earth. Having new and improved methods to optimise supply chain efficiency at a global level will be hugely beneficial to humanity, reducing the cost of goods, wastage, and carbon emissions, and increasing wealth generation. Supply chain optimisation is a complex, multi-objective problem that involves many interconnected components and interacting decision-makers with diverse, and often conflicting goals.  

The problem is often too complex to optimise at a global level, and instead the parts of the system are decoupled, modelled separately and optimised separately. With our partners, Peak AI Limited, we are interested in building AI systems for holistic supply chain optimisation in this project. In particular, the project aims to investigate whether adaptive AI decision agents could be used to make competent decisions about which actions to take in a supply chain, to get closer to a global optimum. The “adaptive” part of the name refers to the requirement that the agents will have to adapt quickly to sudden changes or shocks in the supply chain, such as a natural disaster, failure of production facilities, or key supplier closure.  

To conduct this research, our methodology will draw on concepts from reinforcement learning, multi-objective optimization, and multi-agent systems, and then be validated using several open-source supply chain datasets (e.g. as available on OpenML) and customer supply chain activity data provided by Peak AI.  

Potential research questions include: 

  • Can we create realistic holistic supply chain models that can be optimized autonomously?  
  • Can the AI decision agents be trained to maximise global “whole-system” objectives to meet long-term environmental, social, and economic value for all stakeholders involved including the public? 
  • How can multiple conflicting objectives best be considered by the decision agents? 
  • Are AI decision agents able to adapt rapidly enough to avoid the need for manual intervention during supply chain shocks? 
  • Which machine learning algorithms and simulation tools are best suited to the requirement of rapid adaptation to new supply chain conditions? 

 

Project deadline: MN61

Application deadline: 1st April 2022

How to apply