Creating Dr. Dom
As a tech startup, Domos has always been heavily focused on R&D. But in the last few months we’ve taken research to the next level by starting an internal PhD program. In collaboration with the University of Oslo and the Norwegian Research Council, I have started a PhD project with the title:
“Hierarchical Reinforcement Learning Models with Applications in Radio Resource Management”.
The purpose of the project is to develop new methods in machine learning to solve problems in the field of management and control of home networks. The main problem in home networks is how to manage scarce radio resources in the unlicensed spectrum. The lack of management of radio resources cause unpredictable latency and throughput for end-users, ultimately degrading Quality of Experience¹. Existing methods are insufficient for managing the radio resources because the amount and diversity of devices have become too big for hand-engineered solutions². I addition to this, solving some problems require a global view of the network which has only recently become available. This opens up new avenues of research. Machine learning provides promising techniques to solve these problems, but more research is required³.
Most research on the networking technology used in home networks rely on simulations because it is difficult for researchers to get access to running experiments in real-world home networks. The project will leverage new cloud-based solutions, already implemented by Domos, which enable data-driven control and management of radio resources. The cloud-based solutions make it possible to empirically test and compare machine learning models and control policies in real world home networks.
: C. Pei, Y. Zhao et al. , “Wifi can be the weakest link of round trip network latency in the wild,” in Proc. IEEE INFOCOM , 2016
: Zhang, C., Patras, P. & Haddadi, Deep Learning in Mobile and Wireless Networking: A Survey, 2018, arXiv:1803.04311, URL: https://arxiv.org/pdf/1803.04311.pdf
: Zheng, Kan & Yang, Zhe & Zhang, Kuan & Chatzimisios, Periklis & Yang, Kan & Xiang, Wei. (2015). Big Data Driven Optimization for Mobile Networks towards 5G. 10.13140/RG.2.1.2389.1923. URL: https://www.researchgate.net/publication/283492290_Big_Data_Driven_Optimization_for_Mobile_Networks_towards_5G