Multi-fidelity modelling for networks of simulation models.

Supply chains are massively complex systems, which depend on successful interactions between thousands of independent elements. For example, a store will order stock of a particular product to anticipate demand from a distribution centre. The distribution centre must know roughly what the stores it services wants, to reduce the delays associated with ordering from the supplier. To be able to improve the performance of a supply chain it is useful to simulate it, so we can test its behaviour under different conditions.
However, an accurate simulation of even the smallest supply chain will have to repeatedly run certain component parts to respond to the feedback from other components up and down the chain.
Therefore, the goal of this project is to develop methods to expedite these simulations for the purposes of experimentation. One way to do this is to run approximate simulations for less relevant parts of the supply chain; to speed up the simulation at the cost of its accuracy. We will develop state-of-the-art computational methods, which dynamically and intelligently decide on simulation complexity levels, in an effort to balance efficiency and accuracy.
The novel research from this project is at the interface of Operational Research, Statistics and AI, with practical applications improving the productivity of retail and industry logistics.