PHILADELPHIA â Researchers at the University of Pennsylvania are participating in a massive, interdisciplinary collaboration known as the TerraSwarm Research Center, which will study the potential applications â and risks â of âswarm-basedâ computing and robotics. Based at the Swarm Lab at the University of California, Berkeley, the nine-university project has received a $27.5 million, five-year grant from the Semiconductor Research Corporation and the Defense Advanced Research Projects Agency as a part of the Focus Center Research Program.
The Penn team, led by George Pappas, the Joseph Moore Professor and chair of the Department of Electrical and Systems Engineering, and Vijay Kumar, the UPS Foundation Professor in the departments of Mechanical Engineering and Applied Mechanics and Computer and Information Science of the School of Engineering and Applied Science, will receive $2.75 Million.
For the TerraSwarm Center, âthe swarmâ is an extension of âthe cloudâ that goes well beyond information technology to improve energy efficiency, safety, comfort, security and human effectiveness by integrating the physical world with the cyber world. It leverages recent advances in the variety, cost, size and power consumption of sensing and actuation devices and the associated communication networks. Sensors on these devices can collect a range of data â including video, audio, location, movement, temperature and air quality â that can be used by computing systems to direct physical devices in smart buildings, transportation systems, medical systems, security systems and homes.
Sensor and actuator swarms were originally conceived and developed in the mid-1990s, but thus far they have primarily been used in the context of single, localized applications, such as monitoring and controlling the temperature in commercial data centers. A key goal of the TerraSwarm Research Center is to enable swarms to be deployed at a much larger scale, extending across a city and beyond, and to interact synergistically with each other, with pervasive networked handheld devices and with the computing "cloud." Swarm data can then be combined with information already available from cell phones and social-networking applications, dramatically increasing the potential for sophisticated data analysis and correlation.
âIn emergency situations,â Pappas said, âlike responding to 9-1-1 calls or surveying the environment after a disaster like Hurricane Sandy, there is an immediate need for rapid and reliable collection of information, which might be done by stationary sensors on the ground or by mobile sensors on robots. Information could then be integrated from various sources in cloud-based services and relayed back to the ground for optimal deployment of search-and-rescue teams.â
There are risks involved when it comes to collecting information on such a massive scope.
"Connecting sensors and actuators to the cloud is like giving our cyber world eyes, ears, hands and feet," said Edward A. Lee, the center's director. "We can use these capabilities to provide large-scale services like better traffic control, energy efficiency and emergency response, not to mention improvements in quality of life. But these services must come with assurances of safety, security and privacy, a far-from-trivial challenge."
âThe open swarm platform concept, proposed by the TerraSwarm Research Center,â said Jan Rabaey, the centerâs associate director, âopens the door for a wave of unparalleled creativity and innovation, leading most likely to applications and functions we cannot even predict today.â
The range of possible applications is stunning, but so are the challenges. Of significant concern are data privacy and security. If the cloud is able to affect the physical world, it is essential that it withstand malicious tampering and sensor failures. It is also critical that the data collected is not used in malicious ways.
âEmerging systems,â Pappas said, âsuch as the Smart Grid or intelligent transportation systems often require end-user applications to continuously send information to the cloud, which performs monitoring or control tasks. For example, including smart meters in your house may improve energy efficiency but also enable the energy company to infer what you are doing while you are home.
âA major challenge that the Penn team will be working on is balancing the utility/privacy tradeoff where users send enough information to benefit from such services but not so much information that private activities can be uniquely identified.â
A further challenge is that TerraSwarm applications must be able to dynamically identify and recruit local resources â such as sensors, mobile display screens, communication channels or even unmanned vehicles â to respond to service requests. This capability will require significant advancements in operating systems and resource-allocation algorithms.
âEstablishing and maintaining communications capabilities, between humans and the cloud as well as among members of the mobile network is critical,â Kumar said. âWe will be working on algorithms and architectures for mobile swarms that allow different kinds of units to function as a cohesive group. We will be using quadrotors designed by KMel Robotics, a GRASP Laboratory spinoff, to create mobile swarms and experiment with the algorithms and architectures.â
To address the multi-disciplinary challenges associated with TerraSwarm applications, the team includes 20 engineering faculty members at nine research universities. In addition to Penn and Berkeley, partner institutions include the University of California, San Diego; the University of Illinois; the University of Washington; the University of Texas at Dallas; Carnegie Mellon University; the California Institute of Technology; and the University of Michigan.
Experts in data security, sensors, actuators, operating systems, development tools, robotics, energy efficiency and communications will use the TerraSwarm Research Center as a home base for sharing ideas, developing applications and developing an open extensible platform that can unleash the creativity of millions of potential âswarm appâ developers.