In my research I am investigating how machine learning and big data analytics can drive the development of intelligent communication networks. I believe the advent of high throughput, ultra-low latency applications such as autonomous vehicle communications and virtual reality will make autonomous, self-adaptive networks paramount to enabling any scalable deployments of 5G services at sustainable costs. The upcoming era of 5G networks is therefore an exciting opportunity to conduct research that lies at the intersection of ubiquitous connectivity, intelligence, and distributed systems.
1. Machine Intelligence and Big Data Analytics for 5G Networks and Beyond
NSERC Strategic, Queen’s University – Ericsson
Overview: Topics of investigation include network traffic prediction using generative artificial intelligence models (for IoT, augmented reality, autonomous vehicles traffic), hybrid deep learning architectures for spatiotemporal mobility modeling, anticipatory edge computing and NFV using reinforcement learning, predictive mobility-aware scheduling.
2. Autonomous Multi-RAT Radio Resource Management for 5G
OCE ENCQOR 5G Academic Technology Development Program
Overview: The goal of this project is to investigate how unsupervised machine learning and deep reinforcement learning techniques can be used in dual connectivity management of 5G networks. The objective is to autonomously adapt control plane and data plane operations based on contextual network and environmental data.
3. Deep Learning for and Signal Strength and Mobility Prediction (more details below)
Overview: Developing deep learning including recurrent neural networks to predict signal strength and user location.
4. Network Architectures Beyond 5G for Autonomous Vehicle Communication
Overview: Investigating predictive architectures for scalable uplink video delivery of autonomous vehicle data traffic.
Predicting the Radio Signal Strength along Public Transportation Routes
In this project a database of signal strength measurements and GPS coordinates is generated for a popular bus transportation route in Kingston, Ontario (shown to the left), that covers sub-urban and urban neighborhoods. The data will be analyzed to determine:
Signal strength predictability in different geographical locations and times of the day. Preliminary results indicate that signal strength patterns are repeated along the route (as shown in the measurements below).
Location predictability of the bus at different times of the day. This is to investigate the effects of traffic lights, and congestion on the prediction accuracy as shown below.
Correlation between signal strength and received data rate in Mbit/s. Practical models for data rate variability will then be developed for use in robust resource allocation.
The end goal of the project is to develop prediction algorithms and probabilistic models to enable the practical implementation of predictive resource allocation and content delivery.
Location and signal strength variations with time along the bus route (1 hour) for different bus trips at two times of the day. The variability is more at 12 pm.
Modeling the variation of transport block (TB) size for different signal strength variances. The average TB sizes are (a) 309 bytes, (b) 967 bytes, and (c) 1620 bytes.