Electrical, Computer, and Software Engineering

Applications for 2025-2026 open on 1 July 2025

Reinforcement Learning based Navigation for Autonomous Drones

Project code: ENG020

Supervisor(s):

Henry Williams

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

This research project explores the application of reinforcement learning (RL) to enhance the autonomous navigation capabilities of drones in real-life contexts. Reinforcement learning is a type of machine learning where an agent learns to make decisions by receiving rewards or penalties. It offers a promising approach for developing adaptive and efficient navigation strategies.

RL enables the drone to learn complex navigation behaviours through its interactions with the environment, allowing it to adapt to new conditions. In theory, by leveraging RL algorithms, drones can be trained to navigate complex environments, avoid obstacles, and optimise flight paths or interactions with the environment (in aerial manipulation cases) beyond conventional control methodologies.

The role

This project will focus on designing and testing a real-world environment for training RL algorithms for drone navigation and aerial manipulation/interaction, emphasising sensor data integration and real-time processing capabilities.

Simulation training cannot teach effective real-world behaviours as modelling complex dynamic environments and the dynamic phenomena of task execution effectively are infeasible (conventional control would be applicable otherwise). This leads to the simulation-to-reality gap as agents overfit the simulated and physical environment differences. The platform will be built and tested in 405 on the eighth floor in the Drone lab.

This work will build on the wider work in the Robot Learning Team within CARES (https://cares.blogs.auckland.ac.nz/) which has deployed RL for a range of Roboitcs tasks: https://github.com/UoA-CARES/cares_reinforcement_learning including dexterous manipulation (https://github.com/UoA-CARES/gripper_gym) and autonomous racing (https://github.com/UoA-CARES/autonomous_f1tenth).

Design and formal verification of autonomous vehicle platoons

Project code: ENG021

Supervisor(s):

Avinash Malik

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

In this project the student will design a spiking neural network (SNN)
based autonomous driving algorithm for vehicle platoons.

The developed autonomous algorithms would then be formally verified
for correctness. The implementation and verification of the autonomous
driving algorithms would be carried out using tools developed in the
Precision-Timed machines (PRET) group department of electrical and
computer engineering.

New SNN training techniques using back-propagation, rather than
reinforcement learning (RL), would be investigated in this project. A
new programming language (TimeTide) developed by the PRET group (along
with a leading tech company) would be investigated for implementation
and description/implementation of formal verification.

Functional safety and timing properties of the system would be
verified during the project. The project would get inputs from leading
autonomous vehicle manufacturer.

Success of the project would help in demonstrating: (1) new techniques
for training SNN using back-propagation outpacing the current slow and
error prone RL based learning, (2) value addition in the development
and deployment of safety critical systems using Timetide.

A compact ancillary supply for cryogenic power converter

Project code: ENG022

Supervisor(s):

Duleepa J Thrimawithana

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Cryogenic electronics, also known as cold electronics, is an emerging field experiencing rapid global growth. Its relevance spans multiple high-impact sectors including space technologies, fusion energy, electric aviation, particle accelerators, AI datacenters, MRI systems, and quantum computing. As these sectors advance, cryogenic electronics is expected to become a core component of future high-tech systems.

The role

To-date, there are no records of a fully cryogenic power converter. The team at the University of Auckland is one the leading teams in this area, and have developed the techonlogies that will enable a fully cryogenic power converter. The goal of this project is to miniaturise the ancillary supply that was designed by the group, and using this demonstrate a functional fully cryogenic power converter.

Low frequency wireless power transfer in cryogenic conditions

Project code: ENG023

Supervisor(s):

Duleepa J Thrimawithana

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Cryogenic electronics, also known as cold electronics, is an emerging field experiencing rapid global growth. Its relevance spans multiple high-impact sectors including space technologies, fusion energy, electric aviation, particle accelerators, AI datacenters, MRI systems, and quantum computing. As these sectors advance, cryogenic electronics is expected to become a core component of future high-tech systems.

The role

To-date there is very little work done around wireless power transfer in cryogenic conditions using super conductors. The goal of this project is to explore the possiblity of transferring wireless power from room temperature to cryogenic temperatures and the feasibility of using super conducting tape to improve system efficiency.

Power electronics for hydrogen electrolysers

Project code: ENG024

Supervisor(s):

Seho Kim

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Hydrogen electrolysers use electrical energy to split water into hydrogen and oxygen through electrolysis, and among the various types, Proton Exchange Membrane (PEM) electrolysers are of particular interest due to their high efficiency, compact design, and rapid dynamic response. These attributes make them well suited for coupling with variable renewable energy sources like wind and solar.

PEM electrolysers require a stable and regulated DC power supply, and power electronics play a vital role in converting and conditioning the electrical input to meet these needs. Using components such as rectifiers and DC to DC converters, power electronic systems convert AC power from the grid or renewables into high quality DC while managing voltage, current, and system protection. They must also respond quickly to power fluctuations and support dynamic operation to ensure efficient hydrogen production and protect the longevity of the electrolyser stack.

The role

This project would work on designing a low-power power electronics circuit used to power a small PEM electrolyser with novel energisation methods to improve the conversion efficiency.

Digital educational engineering

Project code: ENG025

Supervisor(s):

Nasser Giacaman

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Digital Educational Engineering (DEE) is about using an engineering approach to design, build, and evaluate a software-based solution that will address some education-based problem.

The role

This project will involve developing a software application or software tool, which can utilise a range of digital technologies. The particular project selected will be determined closer to the time, after meeting the allocated student, in order to carry out a project that they are technically confident in.

Ideal student

To be successful in this project, you should be a strong programmer confident in using technologies such as HTML, CSS, JavaScript, React, Python, etc.

Beyond Backpropagation: Investigating Local Learning and Backpropagation-Free  Algorithms

Project code: ENG026

Supervisor(s):

Waleed Abdulla

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Backpropagation (BP) is the standard training method in deep learning, but it suffers from high memory demands, limited parallelism, and biological implausibility. Recently, Mono-Forward (MF) and FAUST algorithms have emerged, offering local learning updates and more biologically plausible approaches. These methods promise improved memory efficiency and parallelizability. Other backpropagation-free algorithms can also be investigated.

Objectives:

  • Understand the theory behind BP, MF, and FAUST.
  • Implement MF and FAUST algorithms using frameworks like Google Colab, PyTorch, or TensorFlow.
  • Compare their performance to BP in terms of accuracy, memory usage, and convergence speed.
  • Explore potential enhancements and applications to real-world datasets (MNIST, Fashion-MNIST, CIFAR-10, and larger datasets).
  • Summarize findings in a final report and presentation.

Methodology:

  • Literature Review: Learn about BP, MF, and FAUST.
  • Implementation: Develop baseline BP models, MF models with projection matrices, and FAUST models with similarity-based loss.
  • Evaluation: Test models on benchmark datasets, analyze performance (accuracy, memory, convergence).
  • Reporting: Summarize findings and insights.

Expected Outcomes:

  • A clear comparison of BP, MF, and FAUST models.
  • Source code for reproducibility.
  • Final report and presentation highlighting insights and practical trade-offs.

Deliverables:

  • Source code repository
  • Final report
  • Presentation

VeinPrint: Biometric Data Collection and Matching Using Palm and Wrist Vein Patterns

Project code: ENG027

Supervisor(s):

Waleed Abdulla

Felix Marattukalam

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

This project aims to design and implement a system for the collection of palm vein and wrist vein images to support the optimization of biometric matching algorithms. Vein pattern recognition is a highly secure and contactless biometric modality that leverages the unique subcutaneous vascular patterns of individuals, which are difficult to replicate or forge. These patterns are typically captured using near-infrared (NIR) imaging, which reveals vein structures invisible to the naked eye.

The role

The project involves the creation of a robust data collection pipeline using NIR imaging hardware, potentially adapted for use with mobile or low-cost imaging devices. Collected vein images will be used to train and evaluate advanced machine learning and deep learning-based biometric matching models.

Key goals include improving image acquisition quality, enhancing feature extraction techniques, and optimizing matching accuracy under varying conditions such as lighting, skin tone, and hand positioning.

Tasks

As part of the project, students will collect and annotate a dataset of palm and/or wrist vein images, explore image pre-processing and enhancement methods, and obtain results using the existing biometric recognition pipeline.

The final outcome will include results based on reliable vein pattern recognition and matching, along with recommendations for real-world deployment in identity verification systems.

Energy-Efficient Hardware Solutions for Real-Time Machine Learning

Project code: ENG028

Supervisor(s):

Maryam Hemmati

Morteza Biglari-Abhari

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Rapid advancements in Machine Learning (ML) have resulted in widespread deployment of ML solutions in several areas. Recent advances in semiconductor device technology and hardware architectures, data processing, and computing are shifting ML solutions towards the edge, close to data sources.

Deploying state-of-the-art ML algorithms requires high-performance yet low-power hardware architectures. Heterogeneous computing platforms are introduced to optimise performance and energy efficiency. Several ML applications require to meet real-time constraints as well.

The role

This project aims to investigate performance and energy-efficiency advantages provided by AMD AI engine technology and develop customised ML hardware accelerators on heterogeneous Versal platforms to improve both energy consumption and performance for real-time applications.

Requirements

This project requires a strong background in digital system design and computer system architecture. Competency with Electronic Design Automation (EDA) tools is required. Applicants should be willing to learn to work with new design tools from AMD.

Synchonous Language Control of an Electric Vehicle

Project code: ENG029

Supervisor(s):

Matthew Pearce

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

This project fuses synchronous distributed computing and power electronics with the goal of improving control and communication in Electric Vehicles.

The University of Auckland is world leading in the field of cryogenic power electronics. Currently in development are motor / dyno systems which are representatives of aviation electric machines, in the scale of drones of a few hundred watts to light aircraft in the 10s of kW. We work closely with Robinson Research Institute at Victoria University of Wellington who are developing designs for multi-megawatt superconducting electric machines to go along side their 10kW superconducting prototype.

The role

To this end we wish to synchronously control up to four electric machines connected to representative dynomometers running a mission profile. This project will involve running the new language 'timetide' which provides a mechanism to describe logical time abstraction and how distributed processes may be synchronized within it.  

Autonomous mobile buoy for monitoring the health of New Zealand lakes

Project code: ENG030

Supervisor(s):

Kevin Wang

Akshat Bisht

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

The water pollution in New Zealand lakes is caused by the runoff from extensive farming in the lakes' catchment areas. Efforts are now being undertaken to improve the quality of the lakes, and it is important to have technologies to continuously monitor the water quality.

The role

In this project, we aim to develop a mobile buoy which can carry a set of water quality sensors and allow monitoring of the lake’s water quality at multiple points around the lake. The craft will be autonomous and capable of reporting data continuously to a cloud-based backend.

The project will explore the various technologies such as Low Earth Orbit Satellite communication to achieve remote control, autonomous navigation, sensing, and communication, which the buoy will use to cover the entire lake collecting water quality data.

Test-Driven Development for LLM-based Code Generation

Project code: ENG031

Supervisor(s):

James Tizard

Valerio Terragni

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Large Language Models (LLMs) have shown remarkable capabilities in generating software code from natural language prompts provided by humans. As these capabilities rapidly evolve, software development is increasingly moving toward a human-directed approach, where developers primarily focus on specifying requirements for implementation by LLMs.

This emerging paradigm raises important questions about the evolving role of requirements engineering. One promising avenue is Test-Driven Development (TDD), where software requirements are initially defined as test cases, guiding the LLM to produce code that meets these tests.

The role

Preliminary studies suggest that applying TDD techniques enhances the accuracy and reliability of code generated by LLMs. However, numerous open questions remain. For instance, can natural language requirements effectively drive the automated generation of test cases by LLMs? Additionally, which testing methodologies and coverage criteria are most effective in guiding the synthesis of accurate, robust code by LLMs?

This project offers a practical opportunity to directly engage with these emerging questions by applying state-of-the-art LLMs to real-world natural language requirements.

Skills gained

You will gain hands-on experience with techniques for guiding and evaluating generated code, assessing its quality and reliability in realistic software engineering scenarios.

SSIF Future Architecture of the Network: Growing Expertise in the Electrical Industry

Project code: ENG032

Supervisor(s):

Nirmal Nair

TIpene Merritt

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

We will examine the impacts of fragile electricity networks upon rural Māori communities.

Ideal student

Year 3 and above Electrical Engineering students. Students with an interest in applied research.  

Vision Mātauranga (VM)

Mātauranga Māori in Electrical Power Systems Engineering

Project code: ENG033

Supervisor(s):

Nirmal Nair

Deidre Brown

Tipene Merritt

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Matauranga Māori can influence engineering by providing an alternative viewpoint and value system for consideration during initial planning and design such that projects are in line with the local sustainability goals.

The role

The goal of this research is to investigate and current and past works on Mātauranga Māori and how it can influence / interface with power engineering project design towards a carbon neutral and sustainable future. Currently there are no clear guidelines and interface, making conflict resolution between power companies and iwi challenging if a situation arises.

Ideal student

Ideal for any year 3 or year 4 engineering student at UoA.

Mātauranga Māori in Power Engineering - Achieving sustainability and zero carbon futures with countries' indigenous knowledge in design | IEEE Conference Publication

AI-Driven Software Engineering

Project code: ENG034

Supervisor(s):

Valerio Terragni

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

Software engineering is changing. AI systems, especially Large Language Models (LLMs) like ChatGPT, are starting to play a big role in helping developers write, test, and understand code. This trend will continue in the future, where AI and human developers will work closely together.

The role

In this project, you will explore how to build tools that use LLMs to support software engineering tasks. You will design and implement a prototype AI-driven tool for software engineering and evaluate how well it works. The final goal is to write a scientific report about your findings.

This project is part of a bigger research effort to understand how AI can be used to improve the way we build software. It’s a great chance to learn how to do hands-on research while working on exciting new technology (LLMs).

Tasks:

  • Build a software tool that uses an LLM (like ChatGPT).
  • Evaluate the tool to see how well it work.
  • Write a report or paper about what you find.


Requirements:

  • Proficiency in programming (e.g., Python or Java)
  • Interest in AI and how it can help software developers

Emotion decision engine based on text using Reinforcement learning and DNN

Project code: ENG044

Supervisor(s):

Ho Seok Ahn

Bruce MacDonald

Discipline(s):

Electrical, Computer, and Software Engineering

Project 

This project will decide emotions from text, which is for TTS. Find the relation between emotion and text, and train it using reinforcement learning and/or Deep Neural Network (DNN). So when the text is given, robot generate its emotion from the given text, and generate emotional expressions. This project is related to ongoing 5 years research project, SHRI, and will work with PhD and professional staff. All necessary devices will be provided.

This project includes to:

  • find the relation between emotion and text, 
  • train the findings and make a model,
  • generate emotional expression based on the emotion, and
  • update the emotion model from user feedback

Project scope will be decided after the meeting with supervisors. You will bring to the role a passion for research and engineering, excellent computing skills (including a high level of programming ability), and a strong sense of responsibility. It will be better if you have experience of vision processing and/or using DNN.