Traffic Estimation and Stable Resource Allocation Using Distributed Machine Learning

The proposed research will develop novel distributed machine learning techniques for stable resource allocation and improving traffic estimation in networks. It is a well-known fact that networks are becoming complex and user demand is growing in many directions including the traditional demand for capacity and less delay, as well as improvements in Quality of Experience (QoE). Backhauling the multiplexed demand over the core networks calls for accurate traffic estimation. On the other hand, control of the resource allocation, based on such predictions, needs stable and robust solutions. This is a highly challenging problem since when multiple agents use the information collected from the field, they may converge to conflicting decisions which risks the stability of the network. In certain cases, the agents themselves might not even converge, let alone the whole network. Therefore, techniques that consider control, stability and assurance must be developed. This project will develop fundamental solutions for next generation ultra-agile networks.

Intern: 
Shahram Mollahasani;Mohammad Sadeghi
Faculty Supervisor: 
Melike Erol-Kantarci
Province: 
Ontario
Partner University: 
Program: