Digital Twin

Overview

• Simulations of urban environments (digital twins) are actively being developed to aid local governments in making decisions about urban development. The decision space – the ways in which urban planners can choose to modify the urban landscape – is too large to search exhaustively (i.e. to simulate every possible development action). A promising possibility is to cast the problem as a multi-objective optimisation, and design an algorithm to solve it.​

​

Detail

Background

• There is growing interest in the use of digital twins for urban planning.

• Such simulations must define a “grammar” with which to describe urban configurations. They then provide predictions of indicator values (e.g. air pollution levels or green space accessibility) for a variable time horizon, given an urban configuration.

• Urban planners target a trade-off between indicators, and make decisions about urban development commensurate with this trade-off.

• Aioi Nissay Dowa Insurance, being a large insurer, has incentives to reduce risk for urban populations as this risk is often covered by our industry. We, therefore, have an interest in empowering urban planners with better tools to make our cities safer and more efficient.

​

Hypothesis

• The existing literature on optimisation for urban planning defines the search space by manually separating urban areas into cells which can be labelled from a set of pre-specified categories.

• This project will instead work with the recently proposed “spatial signatures” framework. This takes a data-driven (as well as theory-informed) approach to specifying spatial units from satellite imagery (among other data sources). A key advantage is that the granularity of the spatial units in this framework can be adjusted, which is necessary to make the optimisation problem tractable.

 

R&D

• Develop a multi-objective optimisation algorithm using a search space defined with spatial signatures to target wellbeing indictors predicted with a land-use digital twin.

• Undertake a proof-of-concept implementation of the algorithm.

​

Next step

• The project is currently in its proof-of-concept phase. Moving forward, we will transfer the digital twin and multi-objective optimisation algorithm to the Japanese context.