Stochastic Characterisation of Delay Reliability in 5G URLLC
(PhD research, Industry collaboration with Siemens AG)
This project analyses latency reliability in industrial 5G URLLC systems by characterising the Delay Violation Probability (DVP). The work studies how often strict latency deadlines are violated as a function of resource utilisation and retransmission mechanisms. A key focus is on HARQ, including decoding and feedback latencies, and on linking protocol-level design choices to end-to-end reliability guarantees.
Hierarchical Inference Learning for Edge Intelligence
(PhD research)
Tiny machine learning models deployed on resource-constrained edge devices cannot process all inputs locally. This project develops online learning algorithms that decide, in real time, which inputs should be processed locally and which should be offloaded to a more powerful remote model. The resulting Hierarchical Inference framework learns an adaptive complexity threshold, achieves provably sub-linear regret, and forms the basis for later work on cost-sensitive event detection with asymmetric misclassification penalties in edge systems.
Optimal Sampling for Human-in-the-Loop Edge Computing
(PhD research)
Human-in-the-loop systems are often oversampled, wasting energy, or undersampled, degrading user experience. This project studies wearable cognitive assistant applications and derives optimal sampling strategies that balance responsiveness and energy efficiency. Both periodic and aperiodic sampling policies are developed, with analytical guarantees under latency constraints dictated by human interaction timescales.
Pareto-Optimal Downlink Beamforming in Multi-Band MISO Systems
(Master’s Thesis)
This work develops analytical and iterative algorithms to characterise the Pareto boundary of achievable rates in multi-cell, multi-band MISO interference channels. Optimal beamforming vectors are derived to maximise sum-rate while exposing trade-offs between users and frequency bands.
Compressive Sampling of Analog Signals
(Bachelor’s Project)
This project studies sub-Nyquist sampling techniques for sparse analog signals using compressive sensing principles. The work focuses on reconstruction guarantees, sampling conditions, and practical implications for reducing acquisition rates without sacrificing signal fidelity.
Adaptive Delta Modulation: End-to-End System Design
(Bachelor’s Project)
A complete hands-on implementation of an Adaptive Delta Modulation system, covering modulation, transmission, and reception over a microwave link. The project highlights real-time system behaviour, noise sensitivity, and adaptation dynamics.
Controller Area Network (CAN): Seminar and Automotive Case Study
A technical seminar and case study on the Controller Area Network protocol, with application to automotive safety systems. The work includes an overview of CAN communication principles and a focused case study on passenger-car ABS systems.
Towards Efficient Distributed Intelligence: Cost-Aware Sensing and Offloading for Inference at the Edge
Optimal downlink MISO beamforming and pareto boundary calculation
Compressive Sampling of Analog signals
Adaptive Delta Modulation - Modulation, Transmission and Reception : A Hands-On
Seminar on Controller Area Network with case study
ExPECA testbed