It might start with the installation of smart light-responsive windows in your office building. Then a drone flight path will be designated above the mall. One morning, you will notice a self-driving car in your neighbor's driveway. Slowly but steadily, the city you live in will become more and more connected, its infrastructure keyed to monitor itself in smart, sustainable ways.
"Smart cities may be a buzzword now," says ECE's Walid Saad, the College of Engineering Faculty Fellow and associate professor. "But in the next few decades, we're going to start seeing it materialize around us."
Saad specializes in the technologies that enable smart cities, including wireless and cellular networks, cyberphysical systems, unmanned aerial vehicles, cybersecurity, smart grids, machine learning, big data analytics, and game theory. As of spring 2018, he is involved in 16 research projects (13 from the National Science Foundation and three from the Department of Defense) that advance various smart city technologies.
Wireless connections for a connected city
"Delivering high-speed, pervasive wireless services to trillions of mobile devices, as anticipated in the upcoming Internet of Things system, requires a huge transformation of today's wireless cellular networks," says Saad. Existing networks were not designed to handle the scale, density, and dynamics of all these devices.
Because there is not enough spectrum available for classical cellular systems, which operate below 6 GHz and handle the majority of wireless device communications, Saad is exploring millimeter-wave communications, using frequencies above 30 GHz. "Nobody is using these frequencies, but they are sensitive to blockage," explains Saad.
Using the standard cell network, we can move our phones around and still get a good signal. With millimeter-wave frequencies, however, just putting a hand between the antenna and base station is enough to disrupt communications. Saad hopes to integrate millimeter-wave communications with the traditional sub-6 GHz communications to take advantage of the benefits of each. Communications would use the millimeter-wave when it is available, and switch to sub-6 GHz when it isn't. This would give users a high data rate with high reliability.
How can we further enhance city connectivity and open these easily-blocked bands to devices? One possibility is the large-scale deployment of drones as flying antenna arrays or mobile base stations.
Saad's team is studying the performance of drones as base stations and the impact of flight time on communication. They are looking at how drones can be effectively deployed to provide ubiquitous connectivity at both sub-6 GHz and millimeter-wave frequencies, how deep learning can be used to intelligently optimize the use of drone base stations, and how game theory can enable self-organizing resource management in drone-assisted cellular networks.
Smart fog computing
Thanks to these fleets of flying base stations, mobile devices in a smart city will have more options about where to send data requests. "Today, we're using cloud computation as the default for all of our device content requests because these huge data centers have high processing power," says Saad. "But there have been problems with wireless links and latency."
Using machine learning algorithms and online optimization techniques, Saad has been exploring smart computing in which individual devices autonomously evaluate options and decide where to send tasks and content requests. Should they use the slower, but more reliable, local "fog" network—or send it to the cloud?
Autonomous vehicles will be among the first technologies to benefit from data processing location answers. Low latency, high reliability communications links are critically important for autonomous vehicles, says Saad.
This is important for all kinds of autonomous vehicles, from self-driving cars to drones that deliver packages.
Cars that communicate with others and the infrastructure cannot afford even a few milliseconds delay between sending and receiving data. "We must jointly design communication and control systems so that the network delivers information fast enough to keep autonomous cars stable, maintain equal distances between each car, and keep consistent speeds," notes Saad. His team is designing control systems for platoons of self-driving vehicles that use cellular networks to boost reliability and meet latency requirements.
Saad's team is investigating how existing cellular infrastructure can sustain autonomous drone users, such as those that will be used for delivery systems (e.g., Amazon prime). Drones have an advantage over cars for wireless communications: they fly high enough that there are no blocks between them and their base stations. They, however, can create interference for users closer to the ground. Using deep learning algorithms, Saad hopes to help these drones plan paths that balance their interference and flight time concerns.
Physical and cyber-physical security
Drones also have unique security risks. "We must consider the security issues that pose risks, both through the network and in the skies," says Saad.
Besides developing machine learning techniques to fortify the systems, Saad and his team are considering problems such as FAA regulations that restrict drones to a maximum altitude of 400 feet, which places them in range of civilian hunting rifles.
This quandary joins countless others where human decisions must be taken into account especially when it comes to system security.
Many times, a network will be well protected against a technologically sophisticated hack, but vulnerable when it comes to something simple or even inane, says Saad. "Optimal defense theories focus on rational players," Saad explains. "Irrational is harder to predict and harder to defend against."
To meet these non-rational attackers, his team is employing game theory, along with the Nobel-prize winning framework of prospect theory, to help plan drone routes that are unlikely to be in the path of attackers. Because drones are sensitive to flight times, an attacker is likely to wait for one along the shortest path a drone could take. Although flight times do matter, Saad points out that dealing with an attacker will take more time than a circuitous path. His team is also working on cyberphysical security problems that relate to autonomous vehicles and smart power systems.
The same considerations translate to resilience in other types of systems. Saad is also working on methods to fortify coastal cities against natural disasters. They are creating computational, mathematical, and simulation frameworks for "anticipatory resilience," meaning the designs will be tailored to and directly installed in the built environment of these cities.
Saad's interdisciplinary work draws from a wide variety of collaborators, including Harpreet Dhillon, ECE assistant professor, and Allen B. MacKenzie, ECE associate professor; researchers from the Virginia Tech Department of Computer Science, the Department of Economics, and the Virginia Tech Transportation Institute; and researchers from Rutgers University, the University of Colorado Boulder, Princeton University, North Carolina State University, and the University of Miami.