Friday, September 28, 2018
11 AM - 12 PM
4G/5G mobile network plays a vital role in offering ubiquitous Internet access for billions of global users and trillions of IoTs, as it is the only large-scale, operational networking infrastructure, in par with the wired Internet, to support wide-area mobility and wireless access. While the ongoing evolution from 3G/4G to 5G has mechanically grown the bulk of system raw capabilities (wider spectrums, higher speed, larger capacity, super processing, etc), we explore a new perspective to innovate 4G/5G networks. We amplify system intelligence by exploiting such capabilities to their full potentials, rather than simply growing raw capabilities. This not only helps to immediately relieve all kinds of nuisances in today's user experience from slow access to lost connectivity, but also make easier to enable tomorrow's demanding apps like immersive AR/VR, tactile Internet healthcare, autonomous vehicles/ drones, and smart X before full-power 5G rolls out and rocks.
In this talk, I will introduce our recent and ongoing efforts to make mobile networks open and intelligent.
Without knowing what goes on and why it happens, it is hard to elevate all aspects of system performance, fault tolerance, and security to the next higher level. However, we encounter three long-standing technical barriers: (1) mobile networks are largely closed. It still performs black-box operations, accessible only to operators and telecom vendors, not friendly to users and researchers. (2) Their software stack is much more complex than the Internet, which is mostly an engineering artifact from decades of efforts and lacks scientific rigor. (3) Control functions are mandated primarily from the infrastructure side and devices act as a rather passive role despite increasing power. We break up all three barriers with a new data-driven, device-centric, software-defined approach. First, we have opened up data-driven mobile network analytics by designing and developing MobileInsight, the first open-source tool to collect full-stack network data at runtime and construct protocol analytics at the phone. Second, we have empowered learning and reasoning functions via domain-specific analytics and inference engines and applied network verification to assert correctness and provable properties. Finally, we have make full use of device capabilities available and opened up an ample solution space of exploring device-based, rather than infrastructure-centric, solutions to 4G/5G. I will present several exciting case studies to demonstrate benefits and promising potentials of mobile network intelligence. We show that new domain-specific applications of AI and data science techniques are deemed essential to the intelligent 5G and open up an unexplored research area.
Chunyi Peng is an Assistant Professor in the Department of Computer Science at Purdue University. Previously, she was an Assistant Professor in the Dept. of Computer Science and Engineering at the Ohio State University between 2013 and 2017 after I received my Ph.D. in Computer Science at University of California, Los Angeles in 2013. Prior to UCLA, she worked as an Associate/Assistant Research at Wireless Networking Group at Microsoft Research Asia. She is a recipient of NSF CAREER award and receives Best Community Paper award at MobiCom'17 and MobiCom'16, CNS'18 Best Paper Award and several best demo awards. Her research interests are on the broad areas of mobile networking, system and security, with a focus on 5G/4G/3G cellular networks, mobile sensing and network security. More information can be found at her homepage: https://www.cs.purdue.edu/homes/chunyi/
Dr. Guan-Hua Tu