By Liz Sheeley
As the world becomes more connected through the Internet of Things (IoT), questions arise about how to use the resulting massive amounts of data to develop smarter policies and procedures. The April 2018 Special Issue of Proceedings of the IEEE focused on Smart Cities and attempted to move the conversation forward about ways and means to build smart cities responsibly. Professor Christos Cassandras (ECE, SE) was invited to be one of three guest editors on the issue; he also co-authored one of the papers along with Professor Ioannis Paschalidis (ECE, BME, SE), who co-authored two.
Cassandras notes that Boston University was one of the first research institutions to focus on smart cities. “As far back as 2007, we envisioned a city as an ‘enterprise’ modeled as a dynamic system and launched some of the first projects for what we eventually called smart cities,” says Cassandras. “This was done under a National Science Foundation research grant with the cooperation of the City of Boston.”
The process for developing this special issue began over a year ago, notes Cassandras. The two other guest editors are Gilles Betis, the founder and chief executive officer of Orbicité, a consulting company dedicated to smart cities and entrepreneurship, and Carlo Alberto Nucci, a professor of electrical, electronic and information engineering at the University of Bologna in Italy. Cassandras noted that a smart city can be defined a number of ways and in this issue they wanted to tackle how big data can be used in real-time to improve processes, policies and functionalities of a city. The topics covered by the issue include data collection and management, energy systems, infrastructures, mobility, transportation, health and social factors, citizen involvement and a collaborative economy.
One paper co-authored by Paschalis and Cassandras, details a way to alleviate traffic congestion without modifying infrastructure. By routing drivers to paths that would benefit overall traffic flow instead of letting drivers choose with individual preferences in mind, peak congestion can be reduced leading to shorter travel times and a reduction in overall congestion.
Using real traffic data, the researchers devised a method to measure the effectiveness of both routing schemes; the first is the scheme under which the system currently operates, a driver-centric scheme, and the second is the new routing scheme which directs drivers to routes that would benefit the overall traffic flow and congestion. They were also able to quantify the shortcomings of the first relative to the second through what is known as the price of anarchy, owing its name to the fact that selfish driving leads to an inefficiency akin to anarchy.
It’s not that some cars or drivers would be given priority to the best routes, rather, the way the best route would be calculated takes current and predicted future traffic conditions into consideration. Imagine seeing two routes pop up before driving to work. One route offers a shorter driving time and uses the highway. The second route is a bit longer and doesn’t use the highway. Most people would choose the shorter route. But if most drivers make similar selections, the highway will get congested and affect all. The new routing system would take that into consideration and seek to spread the traffic throughout the network, thus reducing overall congestion. Many drivers use navigation systems or apps like Waze to find and pick their driving route, so integrating a new system for routes could be relatively simple.
A second paper from Paschalidis is also included in the issue. Instead of transportation, this paper evaluates how electronic health records can be used to predict hospitalizations for patients with chronic conditions. By using the massive amount of data from the records, they developed an alert system for hospitals to pinpoint at-risk patients and then prevent disease progression with treatment, thus avoiding costly hospitalization—electronic data used to increase efficiency and lower the cost of health care. The researchers were able to use health records as the training tool in machine learning for novel algorithms and focused on predicting hospitalizations for patients with diabetes and heart disease. The algorithms were then evaluated for accuracy and interpretability—it was important that the results be accessible to doctors and patients.
The results were algorithms that hospitals could implement into their electronic health records to indicate if a patient is at risk for hospitalization. Each hospital that adopted the practice could tailor the system to only send an alert based on the risk level they wish to operate with; a hospital with a larger budget would most likely adopt an alert system with a lower risk. If the risk factor is set at a low threshold, more patients would hit the threshold, alerting the hospital to reach out and potentially perform preventative care. This way hospitals that are overburdened can still prevent hospitalizations with this algorithm, but only focus on the most at-risk patients. These algorithms help cut down costs because preventative medicine, when necessary, is significantly less expensive than hospitalization.
These two papers, along with the others in the smart cities special issue, provide ways to use the abundant big data to improve the efficiency of many practices. Instead of theorizing if such an algorithm could work or if routing drivers different ways would improve traffic, large data sets allow researchers to actually show that it would. The policy side of smart cities cannot be ignored as privacy has come to the forefront of discussions concerning big data. The editors of this special issue point out that collaboration across disciplines would be the only way to effectively implement any new policy—they call this the smart city paradigm.
Cassandras says, “As we look ahead, we expect the new ideas discussed in the special issue to plant the seeds for exciting research directions leading to realizable novel technologies and for new business models that cities will adopt.”