UrbanTwin: seeing double for sustainability

Digital Twin City © Adobe / EPFL 2022

Digital Twin City © Adobe / EPFL 2022

A consortium of Swiss research institutes, led by EPFL, has begun working on UrbanTwin to make an artificial intelligence driven and ecologically sensitive model of the energy, water and waste systems of the town of Aigle. The aim: to help boost sustainability.

Twins are a fascinating phenomenon: observing how identical twins, even those separated at birth, can resemble each other in appearance, character, ability and personal taste is astounding. It demonstrates the power of DNA, the smallest of building blocks, in creating surprisingly predictable results.

Now, UrbanTwin, a collaboration of Swiss research institutions led by EPFL, plans to make identical twins of another kind, using neural networks instead of DNA to create a double of a Swiss town. Aigle has been chosen due to its size and because it has an extensive range of water sources.

One of ten nationally funded Joint Initiatives addressing the strategic areas of energy, climate and environmental sustainability, Urban Twin aims to develop and validate a holistic tool to support decision-makers in achieving environmental goals, such as the Energy Strategy 2050 and the vision of climate-adaptive “sponge cities”. The tool will be based on a detailed model of critical urban infrastructure, such as energy, water, buildings, and mobility, accurately simulating the evolution of these interlinked infrastructures under various climate scenarios and assessing the effectiveness of climate-change-related actions.

"We want to incorporate information from a range of different sources," explains Jan Kerschgens, Executive Director of EPFL’s Center for Intelligent Systems (CIS), that integrates research in Artificial Intelligence (AI), Machine Learning (ML) and Robotics to foster research streams leading to the development of Intelligent Systems. “We’ll have leading edge sensors, climate science and algorithmic calculations all in one set of tools - a digital twin. The result should be a product which will allow policy makers to make better informed decisions,” he continued.

As a cutting-edge example of what artificial intelligence can achieve in the modern era, UrbanTwin will be a living model, learning from its own performances and growing over time. "We hope it will make possible the optimization of existing resources, while observing climate impact and suggesting how best to tackle these changes," continues Kerschgens. "We are going develop a tag-based system of recommendations to allow the tweaking of a highly complex system.”

"The project was conceived before the energy crisis hit but is important now, more than ever. It would be great to develop a tool that could be easily transposed," adds Kerschgens, "to open this technology to other urban areas in any region of Switzerland."

Technology transfer is a constantly recurring theme, another is inter-institutional collaboration. Five different institutions are taking part: EPFL, ETHZ, WSL (Swiss Federal Institute for Forest, Snow and Landscape Research), EMPA (Swiss Federal Laboratories for Materials Science and Technology) and EAWAG (Swiss Federal Institute of Aquatic Science and Technology).

Killian Wasmer of EMPA has been working on advanced material processing, but increasingly trans-disciplinary work has become a fact of life, “My team has become populated with machine learning specialists and sensor experts because we need to control our use of resources like energy and water using past data.”

From material processing to digital twins might seem a big leap, but this is where tech transfer comes in. “If you take a set of tools being used for the monitoring of a 3D manufacturing process, it is amazing how easy it is to realign them and turn them to managing a town's water management. The input data is different, but the suite of tools can be used in the same way,” explains Wasmer.

The water monitoring system will include selecting the best source of fresh water supplies and measuring its quality, as well as modelling the disposal of waste water. It will include detective work as well, with a system that will track sources of pollution as quickly as possible and send alarms with the origin located and reported.

Giulio Masinelli is a doctoral student, under the joint supervision of Wasmer and David Atienza of EPFL. He has a good appreciation of this project as he is working on a similar approach for advanced manufacturing. “We can generate data by installing sensors on sinks," explains Masinelli, "measuring water quality around the city, the pH level, the salt concentration and other metrics. We will use machine learning to collect observations, and then make predictions - with physical constraints. These constraints are what make a simulation powerful because it becomes a flexible model with lots of parameters.”

“Masses of work goes into applying partial differential equations to the data so that the system can be generalised without a drop in quality coming from physical constraints and unfamiliar data. The result is a neural network which can generate results in a couple of milliseconds: the resolution of the partial differential equations. Then you can fine-tune the parameters so that they will work with all data. You must not stay too close to one dataset if you want good predictions,” he continued.

Peter Bach of EAWAG is another partner bringing vast experience to the table, having completed a PhD and a Research Fellowship in Australia. There, he was working on the modelling of entire water systems, particularly on emerging nature-based solutions to make the systems more sustainable and circular. “We developed planning-support systems in which users can incorporate a variety of different data to explore innovative and sustainable solutions for their cities. In fact, we can actually learn a lot from the gaming industry in terms of graphical user interfaces, as well as information and hardware management. This enables us to build better tools that can interactively support a diverse group of decision-makers.”

UrbanTwin represents a welcome opportunity for these researchers to collaborate with a range of different teams at a difficult time for Swiss scientists. Participation in Horizon Europe was lost to Swiss researchers since the country broke off negociations with the EU in 2021, making national funding the only current option. Wasmer is hopeful that UrbanTwin can repay the investment of the Swiss government, “if we can improve the way city administrators deal with their resources and raise levels of efficiency it would be a really big step.”

Currently, artificial intelligence is used in an ever-increasing number of ways in research, as exemplified by the AI4Science initiative driven for EPFL by the CIS. Kerschgens is convinced that "artificial Intelligence, as it will be implemented through UrbanTwin, promises to provide a great tool to complement decision-makers in their work, searching through vast stores of data to find anomalies, or recommendations, that would take a person too long to find. UrbanTwin will be an artificially intelligent system, and a holistic one: we expect unexpected results.”

Unexpected results should not come as a big surprise here - it's a twin thing.

Authors: John Maxwell, Tanya Petersen

Source: TRACE - Transportation Center

This content is distributed under a Creative Commons CC BY-SA 4.0 license. You may freely reproduce the text, videos and images it contains, provided that you indicate the author’s name and place no restrictions on the subsequent use of the content. If you would like to reproduce an illustration that does not contain the CC BY-SA notice, you must obtain approval from the author.