Faculty Friday: Yannis Paschalidis

Professor Yannis Paschalidis is a professor in the College of Engineering and the Director of the Center for Information and Systems Engineering. He has appointments in the Department of Electrical and Computer Engineering, the Division of Systems Engineering, and the Department of Biomedical Engineering. 

Tell me about yourself.

YP: I’m an engineer by training. I am in Electrical and Computer Engineering, and I have appointments in Systems Engineering, which is a division in the graduate program, and Biomedical Engineering. My background is in information and systems, decision theory, control systems, and data science, with applications in a variety of areas. I would say these days my main application areas are in decision making for robotic systems and healthcare, mostly data science. I do work in some areas of computational biology. More generally I do theory in optimization, stochastic processes, and the like.

In your opinion, what is the most pressing issue facing urban areas?

I think transportation is a big issue, and I think that there is a very strong coupling between the two areas I do work, healthcare and transportation. It’s well known that most of the emissions in a city come from vehicles, and although it’s important to try to reduce overall emissions, I think there are more specific targeted interventions that could be helpful. Not every area is impacted the same way by emissions. The downtown urban areas are impacted more. Areas where low income individuals live tend to be impacted more There is well-known coupling between these types of emissions and specific health problems, like asthma and reproductive health. I think through more careful data-driven analysis it is possible to identify these hotspots of emissions and congestions and make specific interventions that reduce these hotspots and reduce the hell effects of emissions.

Are there any current projects or research you have that you want to highlight?

Both are relevant to cities and urban living, actually. We’re doing quite a bit of work in healthcare, so analyzing data that comes from healthcare facilities. We are collaborating with the Boston Medical Center, the Brigham and Women’s Hospital, and a number of other healthcare facilities in the Boston area. We’re interested, for instance, in predictive modeling, or predicting hospitalizations within a certain period of time. We have worked on hospitalizations that are due to chronic diseases as well as on readmissions, people who are in the hospital, are released, and then show up again within 30 days. This is important for a number of reasons. If you can predict it, then you can try to prevent it, and and preventing hospitalization is much better for the patient if they are treated before their condition becomes so severe that they require hospitalization. Also, from a healthcare system perspective, it is less costly to treat someone in an outpatient setting rather than admitting and treating them in the hospital. We are not only interested in making these predictions, but also coming up with personalized recommendations for individuals and or for physicians as to what are the factors that are driving a specific hospitalization, why are we predicating that someone will be hospitalized, and what can be done to prevent that hospitalization.

We are also doing work in transportation systems. We are working with data made available to us through the City of Boston and through the transportation planning organization of Massachusetts. We have a number of different data sets – some work we did around two or three years ago involved an app that was developed by the city of Boston. The way the app works is that you can turn it on when you are getting in your car and driving, and it uses the accelerometer in your phone to sense bumps in the road and then transmits that information back to the city. They also have the location of where the bump occurred, so the purpose of collecting this information is to identify the bumps that are more frequently reported and then go and fix these issues, rather than having the crews survey the roads and find potholes on their own. This is a so-called “crowdsourcing” way of getting that information. We worked with them in trying to analyze this information. A lot of times, the city would get signals from bumps they couldn’t do much about. What we did was develop an algorithm that takes in the data for every bump and then classifies them into potholes, maholes, etc, so they can better plan how to deploy their crews and what to prioritize fixing. We provided the software to them. It’s an example of a data science/data analytics application. We actually published a paper with the City of Boston along with people involved in the effort, and other research groups are trying to implement the systems in other municipalities.

Another example: we were given a data set with traffic information in eastern Massachusetts for a period of two years. This was very detailed information: speed of vehicles in all streets on a minute by minute basis, so it’s a huge data set.

The question we asked was: what is the price of anarchy of driving in the city? When you drive, you’re either using a map or making a decision on where to go using your knowledge of the streets and your anticipation of the congestion. You may select a secondary road to avoid certain congestion. In your mind, you have a function that scores different streets based on the amount of time you’re going to spend there. You’re making decisions in order to minimize your delay from getting from point A to point B. Everyone is trying to optimize for themselves, and when we all do that, the network converges to an equilibrium. An equilibrium that means traffic has converged to the state that no driver has the incentive to deviate from what they are doing, since they have selected the best route for themselves. This is not necessarily the best for the transportation network. When we make decisions individually, we might both choose road A, and there would be more congestion on road A and no congestion on road B. If somehow drivers were able to coordinate, then you would achieve less congestion. And this ratio of centralized decisions to individual decisions is called the price of anarchy. It’s possible to go coordinate this without communication between drivers, and induce a more socially optimal routing.

So you’re essentially engineering the solutions to urban issues. What then brought you to the Initiative on Cities?

I think [the IOC] has a great network with individuals that are working at the city and state level who are involved in these types of research projects. Maintaining or strengthening our connection to these individuals increases the number of problems that are being presented to us. There are several of us in engineering who are interested in new problems, new applications, and new datasets where we can apply our expertise and our experience and make a difference in the city.

What is your favorite thing about Boston?

What I like about Boston is that it’s a beautiful but also very intellectual city. It’s a population of people who are in academia, in hospitals, in research institutions, and you have an opportunity to interact with all of these people.

What is your favorite city?

Athens and Rome. I think they are both ancient cities, but also very modern. You can walk past the Acropolis or the Coliseum and you have the ancient world living with the modern world.