ECE professor Radhika Mittal has received a prestigious NSF CAREER award granted to early career researchers. She aims to improve 5G cellular connectivity by allocating network resources more efficiently to different groups – or “slices” – of users.
Most people in the U.S. use mobile data from cellular networks every day, whether it’s watching videos, streaming music or gaming. Each generation of cellular networks has to meet greater demands for bandwidth, as users become more bandwidth-hungry with higher resolution videos and streaming content in real time.
In addition, emerging applications such as augmented virtual reality and tele-robotics can potentially revolutionize education, agriculture and healthcare. However, network limitations are still a key hurdle. At The Grainger College of Engineering, Mittal is exploring ways to enable our wireless networks to meet the bandwidth demands of these new applications.
“5G promises to meet the requirements of these emerging applications, but the mechanisms to enable this are still being fleshed out,” Mittal explains.
Mittal’s research focuses on RAN (radio access network) slicing. Cellular network resources can be divided into “slices” allocated to different user groups. One example of slicing is mobile virtual network operators (MVNOs) like Cricket and Mint Mobile, which each take a share of resources managed by major operators such as AT&T and Verizon.
“An MVNO is one example of a slice, but there are other examples as well: a team of drones doing citywide sensing, or a set of IoT devices in a farm, etc” Mittal adds.
The slicing currently takes place in two stages: network operators divide resources among different slices, then each slice operator can subdivide these further. However, Mittal says this approach ignores an inherent interdependency between these two stages. To tackle this, Mittal is designing new mechanisms that will “embrace this interdependency between these two layers of decision making,” she says. “By doing that, we can improve the utilization of network resources and come closer to meeting the requirements of emerging applications.”
For example, instead of a two-stage process, they can add the slice operator’s logic as a subroutine of the network operator’s logic. “Now when the network operator is dividing its resources among slices, it can issue a query to each slice, asking, if I were to give you this resource, which user will you allocate to that?” Mittal says. Knowing the response to such a query can allow the network operator to make more informed resource allocation decisions across slices that can enhance performance.
In her previous research, Mittal worked on network resource management in other contexts such as wired networks and data centers. Until recently, cellular networks have been treated as a “black box” she says, but these networks are now defined more by open standards in software.
Mittal comments: “It’s the right time to start exploring how some of these network resource management decisions happen in cellular networks, while drawing insights from wired networks. Can we draw parallels between them while identifying the challenges that are unique to the wireless sector?”