Carnegie Mellon University’s Living Edge Lab uses the city of Pittsburgh as a test bed of sorts for exploring edge computing. 

Carnegie Mellon Researchers Tap Edge Computing to Resolve Real-World Challenges

Using cloudlets, researchers keep computing power on location, where data is collected to support beneficial applications.

Carnegie Mellon University’s Living Edge Lab, led by Mahadev Satyanarayanan, is developing an application that would provide up-to-the-minute travel information similar to the Google-owned Waze, which continually provides real-time, location-dependent route details based on user-submitted information. 

Unlike Waze, which requires users to manually input data as they are driving, the Living Edge Lab’s solution wouldn’t require human interaction with a device. This factor alone delivers tremendous value by improving road safety, Satyanarayanan says. 

“This kind of crowdsourced knowledge from drivers about the constantly changing state of roads is valuable. The problem with Waze is that it’s human-driven, and because the vast majority of cars today are occupied by one person, ultimately the input for this source comes from human beings who should be driving,” he says. “Inherently, there is temptation to pick up your phone while you’re driving and report something, so one less distraction is better.” 

The key to enabling autonomous interaction lies with edge computing, which would distribute computing power to areas that use it rather than to a central data center located elsewhere.

Small Data Centers Process Data Closer to the Action

Recognizing the potential to shift paradigms, network and technology developers like the Living Edge Lab are investing in research and development to move computing resources closer to where data is generated to enable a host of new interactive and augmented reality applications, such as the driving details app. For applications that require near-instantaneous response times, cloudlets are preferable to traditional cloud computing, which can suffer from end-to-end latency issues

Rather than being sent to a single centralized data center, data goes to a nearby “cloudlet,” which is essentially a small data center that may consist of a single rack of computers in a closet or a small disk drive in a vehicle that employs multilatency, elasticity and other cloud-computing features

Satyanarayanan and his team are incorporating an edge node called a vehicular cloudlet to work in concert with a video camera to perform continuous video analytics on what the car encounters. One of the challenges they face is determining whether video should be processed within the vehicle or at a nearby zone cloudlet that covers a certain area; both are valid approaches to edge computing, although the in-vehicle processing may cost more. The lab also explores ways to deploy algorithms to conserve bandwidth by suppressing video transmission after three or four vehicles have reported an issue, while allowing zone cloudlets to inquire whether situations have been resolved.

“If you’re not careful with the algorithms used to suppress reports, what happens is that old reports may have changed and need to be revalidated,” Satyanarayanan says. “So the zone cloudlet may probe the vehicle to verify a report of an obstacle from hours earlier.”

MORE FROM STATETECH: Find out how smart cities gain efficiencies from traffic sensors. 

Pittsburgh Test Bed Moves Research from Theory into Practice

As part of the Open Edge Computing Initiative, the Living Edge Lab validates ideas and provides unbiased, critical thinking about which edge applications work and which may need to be revisited or rethought altogether.

The lab uses the city of Pittsburgh, where CMU is located, as a test bed of sorts for exploring edge computing and applications that generate large volumes of data and require intense processing with near-instantaneous response times. Antennas positioned throughout the city are connected via fiber optics to a cloudlet in the lab, which can be tapped into by signals from mobile devices in those areas. 

Researching edge computing theory is important, but nothing can replicate hands-on experience, which is exactly what students and professors receive in the Living Edge Lab, thanks to partnerships with leading technology providers. 

Last year, the Living Edge Lab and Microsoft announced a two-year agreement to drive innovation in edge computing. Under the partnership, Microsoft provides the lab with edge computing hardware and software centered on the company’s Azure offering. Microsoft also supplies Azure credits, which grant the Living Edge Lab access to cloud services like artificial intelligence, the Internet of Things, storage and more. Intel, which already is associated with the lab, also is contributing technology. 

Armed with these technologies, CMU researchers work on a number of projects to bring AI to connected vehicles, drones, factory equipment and other technologies, which allow them to constantly adapt and respond to environmental conditions and changes. Some of the scenarios in which these capabilities are critical include search and rescue, disaster recovery and public safety.

MORE FROM STATETECH: Find out how cities can overcome challenges to smart city deployments. 

Researchers Consider Smart City Applications for Cloudlets

The Living Edge Lab’s vehicle-based solution provides knowledge and insights that have numerous applications for smart cities, most obviously providing real-time reporting on the status of roads to departments and individuals responsible for traffic management and road repair. The introduction of a small, inexpensive, one-terabyte drive to a vehicle expands those possibilities significantly by allowing a vehicle to “remember” everything it has captured — and where — over a month’s time. 

For example, if a crime were committed, a zone cloudlet could contact vehicles in the area to request help in the form of video from particular GPS coordinates at a specific time. This could all be done automatically, with human involvement limited to authorizing the vehicle to share the video. Police are already using GPS data from Google to track potential suspects, as The New York Times recently reported

“Because cars remember the GPS coordinates where they’ve driven in the past month, they can determine whether they might have useful information to share. And then this might be used by public safety and law enforcement to help them with whatever crime they’re trying to solve,” Satyanarayanan says.

Another potential municipal use is for vehicular and zone cloudlets to augment Amber Alerts, which can ping all smartphones in a particular area with messages about child abductions or disappearances. Edge technology and facial-recognition capabilities could extend this to vehicles, offering the potential to fill in gaps between the time of a child’s disappearance and when an Amber Alert goes out. 

“If you send out the child’s photo and you know where the last sighting was, vehicles that may have passed through at that time could report whether they have video of the child on their disks,” Satyanarayanan says. 

“This is very cool research, and we're very excited about this because it's an excellent example of how edge computing can deliver real value,” he says.

benedek/Getty Images
Apr 22 2019

Sponsors