On Digital Twins

This week, I was fortunate to attend a Smart City Digital Twin Convergence Workshop at Georgia Tech, put on in partnership with the National Science Foundation, Stanford University, University of Illinois at Chicago, and Columbia University. Georgia Tech’s Smart Cities and Inclusive Innovation Program under the leadership of managing director Debra Lam put on the workshop focusing on Digital Twins along with John Taylor, the director of the Networks Dynamics Lab also at Georgia Tech.

Practitioners from cities like Pittsburgh, Louisville, Chicago, Columbus, GA, Atlanta, and Syracuse (!!) worked alongside academics from a number of universities, and representatives from the private sector. Though most in the room had some level of expertise in the area of digital twins, it quickly became clear that because this concept is so new, there were differing visions, concepts, and ideas about what a digital twin is or could be. Because the concept is new, and the goal was to figure out how we should be preparing research questions for the future, the setting was less intimidating. Open brainstorming was welcomed.

What is a digital twin?

As mentioned, the concept of a digital twin is very new. The definition according to one of the original papers considering its existence, written by Mohammad & Taylor, 2017 is as follows:

“A Smart City Digital Twin is a smart, IoT-enabled, data-rich virtual platform of a city that can be used to replicate and simulate changes happening in the real city to improve resilience, sustainability, and livability.”

While that definition gives some guidelines, it is so broad that much could fit inside. Generally, I think a digital twin is a digitized representation of the real-world where the data collected in the digital world can be manipulated or modeled to understand what might happen in the real world if a change is made. This is still broad and questions remain.

Is it a digital twin any time an IoT device is used? Do IoT devices need to be deployed everywhere? Does all data need to be collected? Is data that has been collected for a century in cities (like engineering drawings for water mains) included in a digital twin? Should we focus on a specific problem like traffic or try to collect data across the entire city? How do we account for unpredictable things humans do that cannot be easily collected, and are almost impossible to model?

These types of questions were discussed, and the specific definition and its potential for application in a city is still in question, but I thought some of the conclusions that a digital twin could serve to answer short-term questions or very long-term questions and could focus on sudden shocks or broader stresses was interesting — my colleague Grace Simrall from Louisville was behind many of these ideas.

The general sense of the definition was enough to get my wheels turning, thinking both about the potential and challenge of a digital twin.

Where is its potential?

To understand the potential of a digital twin, it might be helpful to consider some applications in the City of Syracuse:

• A couple years ago, we worked with the Data Science for Social Good program to build a model that predicts the risk of water mains breaking. We used historic data about water main materials and previous breaks to build the model. While convening all of this data would likely not be considered a digital twin, more real-time information about leaks, water and soil quality, and corrosion of pipes all converging might be considered a digital twin. If we knew everything about the water mains, we could model where the risk might be highest on a day-to-day basis, and could also model the impact of having one water main break — how will it impact other mains, water usage by residents, etc.
• As the City upgrades its street lights, we hope to understand in real time which lights are on and off. We currently rely on complaints from residents to tell us if a light has gone out. Sensors in the lights can give us better information. Paired with crime information or pedestrian counts on streets, we could build a digital twin that models what happens if we dim or brighten the street lights, or if a light goes out unexpectedly. This way, we could test different scenarios or predict outcomes in a digital world, understand the likely result, and then deploy in the real world. It would reduce the risk of testing in the real world, where the consequences of a bad pilot test could be harmful or annoying to residents.
• For more than a decade, there has been a debate in Central New York about the I-81 viaduct that cuts through the middle of Syracuse. On one side is a group that advocates for the viaduct to be torn down and replaced with a street grid. The argument here is that it activates the space, calms traffic, reduces noise and air pollution, and helps rebuild a neighborhood. On the other side, the group argues that the viaduct should be rebuilt. The argument being that a street grid would make commutes longer and hurt businesses in the suburbs, while also redirecting trucks and other big vehicles to streets that are not built to handle them. A lot of research has been done taking all of these considerations into account, but a digital twin would allow for models to be built predicting impacts for both considerations. It would also potentially show long-term effects on transportation, business growth, etc. The amount of data that would need to be collected and analyzed would be massive, and it is questionable if this is doable, but if it were, decision makers could feel even more confident about their stances on the issue.
• It snows a lot in Syracuse, and especially as climate change shows more impacts across the country, we can expect heavier snow falls and fluctuating weather — freezing and thawing cycles that impact the ability of snowplows to clear the streets and sewers to handle melting snow. A digital twin that measures weather, traffic flows, snowfall, and building occupancy could help to recommend dynamic plowing routes based on where the worst snowfall is and could redirect traffic to only the safest, most recently plowed streets.

What are the challenges?

Most of the examples I listed above are massive in scale when considering how much data would need to be collected and how many sensors would need to be deployed. Aside from there not being a clear definition for a digital twin, there are many challenges:

• There are questions about how data would be collected. Sensors in the field may last for 3–5 years, and when considering blanketing the city in sensors, the costs of hardware can be huge. When batteries in the sensors need to be replaced or hardware needs to be fixed, the ongoing operational costs might be too high.
• There are always opportunity costs, and when cities invest in technology to better maintain infrastructure, there is often a question about why those dollars aren’t just being used to pave a road or fill more potholes. The argument, of course, is that with better data and information we can make more informed decisions about how to best spend limited dollars on infrastructure upgrades or other city projects, but the question will remain a challenge to deploying technology at the scale needed to create a digital twin.
• If digital twins are a representation of what is happening in the real world, then it means a lot of information will be collected about people. Questions about privacy when it comes to using data, sensors, and cameras in government are already challenging. Building a digital twin adds to those questions. Because of the likely cost of deploying a digital twin, public-private partnerships would be a means for funding. While P3s can be useful and important, there are challenges, then, around data ownership and how data will be used.
• Over the past couple of years, ransomware attacks on city governments have increased and damaged the ability to deliver many core services. If access to data in a digital twin were cut off by a ransomware attack, it could further harm a city, especially if the digital twin was helping to guide traffic or respond to a weather issue. Cities need to prepare for cybersecurity challenges in general and also need to have plans for continuity of service even in the case of an attack, but building up a digital twin would be a large commitment that would change how a city chooses to operate, and managing without it could be very difficult.
• As my colleague Kelly Jin, the Chief Analytics Officer in New York City tweeted recently:

Counting things is often very difficult for cities. The amount of data that goes untouched and remains misunderstood, and also has a lot of value that has not yet been seen, means that cities do not necessarily need to invest in building a digital twin in order to find ways to be more efficient of effective. The potential of a digital twin is huge, but in Syracuse, we don’t have a digitized map of where all the stop signs are. Maybe we could skip ahead of manual data collection in this case, but it would be a challenge.

• Procurement is hard for any technical system. Procurement for a digital twin would be a massive challenge.
• Given the infancy of the digital twin concept, partnership between cities and academia would be necessary for research about best practices for implementation to occur. Partnerships are challenging and city practitioners and their politician bosses have different priorities than academia. Those differences can sometimes be challenging to overcome.

Isn’t this the same thing as a smart city?

Many of the discussions of digital twins had me wondering what distinguished this from a smart city. Smart cities also are poorly defined and can potentially mean anything. As my friend Laura Meixell from Allegheny County, who also attended the workshop pointed out, people sometimes consider a better website a smart city, or a marginally better process that involves technology a smart city.

I think a digital twin falls under a smart city umbrella. A smart city might need a digital twin to inform it. A smart city, though, falls under the umbrella of city operations as a whole. It is a natural extension of how a city does its work, not something completely separate. We don’t want to only respond to complaints about broken infrastructure, we want to monitor the infrastructure so that we know immediately when something breaks. In order to do this we need technology to tell us when that thing breaks. The build out of that technology might create a digital twin, and that digital twin might tell us how that thing breaking has ripple effects on the rest of the city. The decision about how to resolve that thing breaking is what ultimately makes a city “smart” because the city fixes the thing faster, or cheaper, or more effectively.

The workshop was fascinating, and I am excited to see how this concept of a digital twin matures. Being so new and potentially vast, this is a great opportunity for partnership between government, private sector, and academia. Though there are challenges and a need for more clarity on the questions we are trying to answer, this pursuit is potentially transformative and the exercise is worthwhile. The potential for data and technology to help improve lives is there if done empathetically and thoughtfully.