Summer Research Scholarships

If you’re a third-year student or above, you could spend the summer supporting a research project in the Faculty of Engineering and earn a valuable scholarship.

Over the summer the Faculty provides several scholarships to students, offering a unique opportunity for students to gain experience in research—with the added bonus of paid work over the summer trimester.

Please find below a list of the Summer Research Scholarship projects that are available in the Faculty of Engineering over the 2021/2022 summer. Before applying please read the Summer Scholars Scheme conditions to ensure that you are eligible to apply.

Students will be selected on the basis of academic merit, expertise in the research area, and recommendations from staff associated with the project.

Each scholarship will have a value of $6,000, paid in four instalments.

The scholar is expected to contribute a minimum of 400 hours to the project between November/December 2021 and March 2022.

To apply for any of the following projects please apply online, noting the corresponding scholarship code.

Applications for the 2021/2022 Summer Research Scholarships will be open from 6 – 20 September 2021.

2021/2022 Engineering and Computer Science Summer Research Scholarships

There are 55 projects available for Engineering and Computer Science students:

  • Integrating Zeek and SDN
  • Automating red teaming
  • Live 360D VR: Panoramic depth video using real-time spherical LiDARs
  • 3D digital twin for New Zealand bird
  • ORCA/Dala: Object capabilities for concurrent applications
  • MOTHRA—dynamic optimisation of dynamic languages
  • aWall—enhancing agile team collaboration through the use of large digital multi-touch cardwalls
  • Genetic programming for explainable dimensionality reduction
  • Genetic programming for job shop scheduling
  • Towards enabling seamless communication in UAV swarms
  • Genetic programming and machine learning for feature selection and classification for quantitative assessment of nutrient content in horticultural products
  • Wiring 42 to the net
  • Novel methods for data-intensive service deployment in multi-cloud
  • Human body communication network
  • Genetic programming for real-world image segmentation
  • A flying base station for a NOMA-based IoT system
  • Multi-output symbolic regression using multi-tree genetic programming
  • Genetic programming for symbolic regression
  • Machine learning for New Zealand fish data analysis
  • Transfer learning for AI planning
  • Controlling IoT devices using smart contracts on Ethereum
  • Tropical cyclone detection
  • Genetic programming for feature extraction to identify irrigated land in satellite images
  • Feature selection for multi-label classification
  • Machine learning for sound field characterization
  • Explainable artificial intelligence using genetic programming
  • An automated approach to evolving scoring module for Cuckoo Sandbox by utilising genetic programming
  • Building interpretable symbolic regression models with genetic programming
  • Machine learning for tree species identification
  • Evolutionary machine learning and data mining
  • Using oil wells for geothermal energy generation
  • Improving renewable energy usage and support for both customers and power line companies
  • Artificial Intelligence based methods for recognising individual kākā
  • A web tool for visualising and performing statistical analysis on Auckland air quality data
  • Bus emissions reduction tool
  • Emissions reduction—national inventory exercise
  • Design, structure and set-up of a local innovation hub
  • Transitioning to a zero carbon public transport fleet
  • Injury reporting for commercial cleaners
  • Develop a volcano rover prototype
  • Improving hatchery algae assessments with artificial intelligence
  • Oyster identification using artificial intelligence
  • Isolation of glycomacropeptides from milk
  • Multi-objective machine learning and applications
  • Software lifecycle management tool
  • Genetic programming with transfer learning for image classification
  • Network resiliency planning tool—upgrade and porting
  • Automated detection of Green Lipped Mussel shellfish opening
  • Image stitching and stabilisation to improve camera AI inference for aquaculture farms
  • Early detection of a marine invasive species on aquaculture farms
  • Mapping the predicted distribution of harmful phytoplankton after forecasted global warming
  • AI based signal classification for cognitive radio systems
  • Procedurally generated location-based VR game maps
  • Automation of photodetection linearity measurement
  • Automation & image analysis of autocollimator
  •  

See the Faculty of Science website for School of Mathematics and Statistics projects.

Internal projects

Integrating Zeek and SDN

Scholarship code 620

Supervisor: Ian Welch

You will work with a state-of-the-art open source intrusion detection system called Zeek that is widely used in industry. The goal of this project is to take an existing integration with a software defined networking system and develop documentation on how to deploy it and VMs/docker images to help further experimentation.

Automating red teaming

Scholarship code 621

Supervisors: Ian Welch

You will use a state-of-the-art cyber security framework for emulating advanced attackers and set up a virtual environment for demonstrating its capabilities. The aim of the work is to document the application programming interface for CALDERA, write a tutorial on how it can be used, and identify potential ways to integrate machine learning.

Live 360D VR: Panoramic depth video using real-time spherical LiDARs

Scholarship code 622

Supervisors: Taehyun Rhee

This project will provide a unique opportunity to evaluate the state-of-the-art LiDAR sensors (Ouster) which capture a wide angle (360-degree) depth information using the live 3D sensing technology, work with Computer Graphics and extended reality (XR) experts to design a future XR platform for immersive 3D live video for various media applications.

3D digital twin for New Zealand bird

Scholarship code 623

Supervisors: Taehyun Rhee with Andrew Chalmers (CMIC)

We are interested in creating high-fidelity 3D digital twins of New Zealand native birds. Birds are culturally important but many of them are endangered or extinct, e.g. the Huia is extinct, none exist to observe other than taxidermy. To reconstruct a digital twin that faithfully represents the birds, this project will study:

  1. Properties of appearance (it's material and geometric structure) as well as the
  2. Motion (how it walks/flies) of the birds, and
  3. Develop a prototype to model a part of the subjects in 3D computer graphics.

ORCA/Dala: Object capabilities for concurrent applications

Scholarship code 624

Supervisor: James Noble

Crypto currencies and blockchains such as Etherium, Libra, and Bitcoin support smart contracts—financial or legal instruments written in code. Unfortunately, any bugs in the code immediately become bugs in the underlying financial instruments, which can cost hundreds of millions of dollars. This project will develop new techniques to verify smart contracts, taking ideas from the Chainmail specification language and applying them at runtime. This project will suit a student interested in smart contracts, dynamic programming languages, and runtime verification, and who is aiming for a career in cryptocurrencies or further research.

MOTHRA—dynamic optimisation of dynamic languages

Scholarship code 625

Supervisor: James Noble

Moth is a VM for dynamic languages, currently the fastest at implementing transient dynamic type checks. This project will develop new techniques to optimise control structures, replacing the special purpose code used in most other approaches. This project will suit a student interested in Java, VMs, and programming languages, and who is aiming for a technical career or further research.

aWall—enhancing agile team collaboration through the use of large digital multi-touch cardwalls

Scholarship code 626

Supervisor: Craig Anslow

aWall is a large multi-touch display application to help support software development teams. It supports features for hosting Agile team meetings. We are developing new features to support different activities as part of Agile processes.

Genetic programming for explainable dimensionality reduction

Scholarship code 627

Supervisor: Andrew Lensen

EXplainable Artificial Intelligence (XAI)—the development of AI algorithms that make decisions which can be understood by humans—is critically important in our modern, data-driven society. In our big data era, unsupervised AI techniques are required to reduce the dimensionality of data so it can be practically analysed. However, such nonlinear dimensionality reduction (NLDR) methods are not explainable. Genetic Programming (GP) has the potential to be a state-of-the-art method for performing explainable NLDR. This project will develop new methods to perform explainable NLDR using unsupervised Genetic Programming (GP) techniques.

Genetic programming for job shop scheduling

Scholarship code 628

Supervisor: Fangfang Zhang with Yi Mei and Mengjie Zhang

Job shop scheduling (JSS) is an important optimisation problem that reflects the practical and challenging issues in real-world scheduling applications such as order picking in warehouses, the manufacturing industry and grid/cloud computing. Genetic programming, as a hyper-heuristic approach, has been successfully and widely used to learn scheduling heuristics for the scheduling problems. We have published a number of good quality papers in this area and have built good resources, e.g., simulation and implementation, which make it easier for new researchers to work on. In this project, the students will use genetic programming for solving the static and dynamic JSS problems, and compare the results with exact methods or manually designed rules. By doing this project, you will get familiar with how to solve important real-world optimisation problems using genetic programming effectively. You will also be skilful in using genetic programming for different types of JSS such as static JSS and dynamic JSS. You will have opportunities to learn important soft skills such as critical thinking, Java/Python programming, professional writing and presentation, and communication skills. Last but not least, we will highly support you to public your work in PBRF quality-assured international conferences or journals. A good background in Artificial Intelligence (COMP307), java programming is preferred.

Towards enabling seamless communication in UAV swarms

Scholarship code 629

Supervisor: Jyoti Sahni

Unmanned Aerial Vehicles (UAVs), usually called drones are being increasingly used for missions classified as dull, dirty and dangerous for humans; such as real-time surveillance, wildfire monitoring and search operations. However, they have certain limitations related to flight hours, processing power and reliability. These constraints limit the usability of present day UAV systems in large scale operations critical to life, property and business.

This project will study the effect of different mobility patterns and environmental obstacles on energy consumption and communication overheads involved in communicating status updates among member UAVs. The outcome of this project will support building a novel energy efficient model to adapt the movement of member UAVs to ensure a coordinated flight. Prospective student will be capable of programming in C++ / Java and Python and be able program using the SDK provided for programmable UAVs.

Genetic programming and machine learning for feature selection and classification for quantitative assessment of nutrient content in horticultural products

Scholarship code 630

Supervisors: Mengjie Zhang and Bing Xue and Qi Chen

Raw horticultural products vary widely in their nutrient composition due to factors such as genetics, cultivation region, and storage conditions. Quantifying the nutrient content of these products is necessary as a means of quality control. With consumers becoming more health conscious the willingness to purchase nutrient-dense products increases substantially. Many bioactive components affect taste, which can also influence individuals’ consumption. Non-destructive, rapid assessment of these bioactive components is essential to provide the nutritional information of horticultural products to consumers, which also maximises profits. However, fast assessment with high accuracy of these bioactive components is challenging. This project aims to develop a new approach to the use of genetic programming (GP) and evolutionary machine learning method for selecting a small number of important features and building interpretable models for for quantitative assessment of nutrient content in New Zealand horticultural products. The focus will be mainly on classification and symbolic regression. A strong background in python/Java programming (COMP261) and in Artificial Intelligence (COMP307) is required. A good background in statistics and optimisation is desired.

Wiring 42 to the net

Scholarship code 631

Supervisor: Marco Servetto

42 is a novel programming language, getting closer to a first public release.

Novel methods for data-intensive service deployment in multi-cloud

Scholarship code 632

Supervisor: Hui Ma

Service-oriented computing (SOC) and cloud computing are becoming popular computing paradigms of developing and deploying software services. A weather forecast service, for example, can be deployed to a cloud which can then be accessed by service users. Multiple clouds are available with different prices and different levels of quality of service (QoS), e.g. response time. To make the highest possible profit, service providers need to decide how to deploy services to multi-clouds so that the costs of providing services are minimized while the QoS are optimized. Unfortunately, it is computational hard to find an optimal deployment of Web services. Therefore, heuristics will be needed to find near-optimal resource allocations for services.

Human body communication network

Scholarship code 633

Supervisors: Alvin Valera

We are proposing the use of the human body as the communication channel itself. There have been successful demonstrations of human body communication, but the problems of link reliability and multiple channel access have not been addressed yet. We therefore seek to investigate these problems in the context of human body communication. The outputs and outcomes of this research will break new grounds in wireless communication, networking, and should lead to the establishment of an interdisciplinary research initiative involving wireless communications, networking, health, and medicine.

Genetic programming for real-world image segmentation

Scholarship code 634

Supervisor: Ying Bi with Bing Xue and Mengjie Zhang

Image segmentation is an important task in computer vision, which has a wide range of applications in many real-world scenarios, such as medical image analysis, biological image analysis, face image analysis, aquaculture and agriculture image analysis. This project focuses on a real-world application that aims to deal with image segmentation using aquaculture image data. The dataset was collected from a mussel farm in Nelson, New Zealand, and well prepared for the image segmentation task. Genetic programming (GP) is an artificial intelligence technique and has achieved promising results in many image-related tasks. The main goal of this project is to develop a new GP approach to performing image segmentation on these real-world images. The project will seek strategic collaborations with Cawthron Research Institute.

This project requires a third or forth-year student with good knowledge in artificial intelligence and machine learning (COMP307/AIML420, or COMP309/AIML421, or AIML426) and strong programming skills in Python. The supervisors will train the student in basic research skills, which will be useful for their further postgraduate research/study. The student will also learn state-of-art knowledge from Evolutionary Computation and Artificial Intelligence Research Group. More information about the supervisors, please see:

A flying base station for a NOMA-based IoT system

Scholarship code 635

Supervisor: Yau Hee KHO

Imagine a remote area where there is no cellular coverage, and critical sensor data gathered from various Internet-of-Things (IoT) sensors need to be transmitted all at once, how do we provide such access? In this project, we aim to create a small-scale IoT system for such harsh and extreme environments, by using an unmanned aerial vehicle (UAV) as the base station (or access point) flying over and hovering above a region of IoT devices to collect (and transmit) data from (and to) the IoT devices. By carefully adjusting the transmit power, we can employ non-orthogonal multiple access (NOMA) technique to increase the overall data rate of the system, rendering it suitable for accommodating many IoT devices that may be trying to transmit and receive data all at the same time. This will facilitate monitoring of vital environmental signs (e.g. those of seismic, volcanic, pests, and predators) in remote, hostile terrains in New Zealand.

Multi-output symbolic regression using multi-tree genetic programming

Scholarship code 636

Supervisors: Baligh Al-Helali, Qi Chen and Bing Xue

Symbolic regression is a crucial task in real-world problems such as ecological modelling and energy forecasting. However, symbolic regression is more challenging when the problem has more than one output that need to be modelled simultaneously. This problem can be solved via genetic programming by evolving multiple models at the same time. This project aims at developing a multi-tree genetic programming method to carry out symbolic regression models for multi-output regression tasks. This method will be designed to minimise the regression error while improving the interpretability of the learned models.

Genetic programming for symbolic regression

Scholarship code 637

Supervisor: Qi Chen with Bing Xue and Mengjie Zhang

Genetic Programming based Symbolic regression (GPSR) discovers the underlying relationship between input and output variables in the data and express the relationship in mathematical models. Compared with traditional regression techniques which optimise the parameters in the predefine model and numerical regression methods which provide “black-box” regression models, GPSR searches for explicit and interpretable models without any assumption on model structure and data distribution. GPSR is a powerful tool and plays a great role in the discovery of knowledge from data in an era of exponential data growth. Despite the many success stories and a high potential for generating interpretable models, the current GPSR methods still have some limitations in generating interpretable models. Firstly, it is still challenging to understand the functional modularity of the regression models in GPSR. Secondly, GPSR models are typically lack of visual interpretability which refers to inspecting a model via plotting model behaviours. This project aims to develop a new GP approach that can deal with the above limitations and improve the interpretability of SR models. A good background in Artificial Intelligence (COMP 307), python/java programming and statistics is preferred. In the project, students can not only learn the state-of-the-art machine learning techniques but also learn how to do research, which is important for their future study, e.g. the Honours program and the PhD program. During this process, students will also learn how to produce a paper and submit it to an international conference.

Machine learning for New Zealand fish data analysis

Scholarship code 638

Supervisors: Bing Xue (Co-supervisors: Mengjie Zhang and Bach Nguyen))

In New Zealand 60% of every fish caught is processed to low-value (NZ$2 kg-1) fish oil and fish meal, usually for inclusion in animal feeds—a terrible waste of one of Earth’s limited resources. A research programme led by the New Zealand Institute for Plant and Food Research, in collaboration with Callaghan Innovation, Victoria University of Wellington, Otago Wellington and Deakin Wellington, is now seeking to change that. The research team is designing a ‘factory of the future’ inspired by ‘Industry 4.0’, which will process every kilogram of fish into high value products. This is an exciting opportunity to be part of a highly collaborative research programme spanning NZ and Australia. This project will investigate a new feature selection or classification method for the provided spectroscopic data and compare with existing methods.

Transfer learning for AI planning

Scholarship code 639

Supervisor: Yi Mei

Transfer learning is a very powerful machine learning technique that leverages the relationship between different problem domains (e.g., images from different sources/websites), and has achieved great success in improving the learning performance in many classification and regression machine learning tasks. General AI requires extracting common knowledge from related problem domains, and transfer the knowledge to solve each problem more effectively and more efficiently. A typical example is AlphaZero by Google DeepMind (an AI game agent that learns to play a variety of games such as Go, Chess, shogi, etc). In contrast with classification and regression, it is more challenging for knowledge extraction and transfer for AI planning. This project will investigate and propose novel transfer learning approaches for AI planning. We will consider different AI planning domains such as games, pathfinding and scheduling, and mainly take the evolutionary computation and learning approaches.

Controlling IoT devices using smart contracts on Ethereum

Scholarship code 640

Supervisor: Winston Seah

This project deals with an emerging technology, the Blockchain along with Internet of Things (IoT). Blockchain already has more than 10 years history yet struggles with its application to IoT environments because of limited scalability and performance. To solve this problem, the first step is to define the architecture accurately and investigate the issue in depth. During the project, you will acquire corresponding knowledge as well as have hands-on chances to develop an application using smart contract on Ethereum.

Tropical cyclone detection

Scholarship code 641

Supervisors: Aaron Chen (and Harith Al-Sahaf)

Tropical cyclones have substantial world-wide social and economic impact. A Tropical cyclone usually starts its life as a tropical disturbance which is marked by a cluster of thunderstorms on a satellite image. The genesis phase of a tropical cyclone beyond a tropical disturbance requires constant monitoring by a weather forecaster. The monitoring process is labour intensive and error prone. At times a disturbance can look to pose little threat then suddenly become more organised into a tropical storm. In this project, we aim to develop innovative evolutionary computer vision and deep learning technologies to facilitate automatic and accurate cyclone monitoring. Through processing a sequence of satellite images and other weather forecasting data, the new computer vision system is expected to accurately detect the formation of tropical cyclone in New Zealand. Through this project and close collaboration with MetService, you will have the unique opportunity to obtain hands-on experience of developing cutting-edge computer vision technologies. You will also deepen your machine learning knowledge and sharpen your research skills by solving a practically important real-world weather forecasting problem.

Feature selection for multi-label classification

Scholarship code 643

Supervisors: Bach Hoai Nguyen and Bing Xue

Nowadays, multi-label classification is getting more popular where a real-world object can be classified to more than one category or label. Feature selection is an essential pre-processing step which can improve the classification performance via selecting complementary features. However, it is not trivial to develop a multi-label feature selection since there are a large number of labels to consider. The goal of this project is to analyse the label relationship, and then decompose the original label set into label subsets which are easier to be addressed by feature selection. This project requires a third or forth year student with good knowledge in machine learning (at least COMP 307) and strong programming skills in Java or Python.

Machine learning for sound field characterization

Scholarship code 644

Supervisor: Bastiaan Kleijn

It is difficult to determine the sound field between microphones, because the number of microphones is generally insufficient in number and the physical context is usually not well specified. Machine learning has the potential to solve this problem because it facilitates a structured way of finding a likely solution from partial information with relatively low computation effort. Transformers are particularly suitable as they allow as input an unordered set. You will implement and test a Transformer based sound field interpolation system. The method will be useful for augmented and virtual reality.

Explainable artificial intelligence using genetic programming

Scholarship code 645

Supervisors: Andrew Lensen and Yi Mei

Commonly used AI algorithms such as Deep Neural Networks (DNNs) are often called "black boxes", as the way in which they make decisions is highly opaque. Genetic Programming (GP) has clear promise for enabling eXplainable AI (XAI), but significant further research is needed to produce understandable, effective, efficient, and trustworthy algorithms. This summer project will investigate novel approaches for applying GP to XAI.

An automated approach to evolving scoring module for cuckoo Sandbox by utilising genetic programming

Scholarship code 646

Supervisor: Harith Al-Sahaf and Ian Welch

Cuckoo sandbox is a tool for advanced analyses of malware/benignware. It uses a scoring module that assigns severity scores to each binary analysed in it. This score shows the malice intent of the respective binary. It has been observed that for some samples of malware, the scoring module produces a score beyond its stated range (i.e. 0 through 10). This is mainly due to its handcrafted scoring mechanism, where it only performs a summation of the severity score for every signature (stored in Cuckoo’s repository) found during the analysis of a binary. The challenge is to evolve a scoring module automatically that can assign a severity score relative to the actual malice intent of each binary analysed in Cuckoo.

Building interpretable symbolic regression models with genetic programming

Scholarship code 647

Supervisors: Qi Chen (and Bing Xue and Mengjie Zhang)

Genetic Programming based Symbolic regression (GPSR) discovers the underlying relationship between input and output variables in the data and express the relationship in mathematical models. Compared with traditional regression techniques which optimise the parameters in the predefine model and numerical regression methods which provide “black-box” regression models, GPSR searches for explicit and interpretable models without any assumption on model structure and data distribution. GPSR is a powerful tool and plays a great role in the discovery of knowledge from data in an era of exponential data growth. Despite the many success stories and a high potential for generating interpretable models, the current GPSR methods still have some limitations in generating interpretable models. Firstly, it is still challenging to understand the functional modularity of the regression models in GPSR. Secondly, GPSR models are typically lack of visual interpretability which refers to inspecting a model via plotting model behaviours. This project aims to develop a new GP approach that can deal with the above limitations and improve the interpretability of SR models. A good background in Artificial Intelligence (COMP307), python/java programming and statistics is preferred. In the project, students can not only learn the state-of-the-art machine learning techniques but also learn how to do research, which is important for their future study, e.g. the Honours program and the PhD program. During this process, students will also learn how to produce a paper and submit it to an international conference.

Machine learning for tree species identification

Scholarship code 648

Supervisors: Bing Xue and Mengjie Zhang

Trees play an important role in the wider ecosystem, providing for example carbon sequestration and regulation of climate, water, and soil erosion. With the increasing availability of remote sensing techniques that can capture above-ground objects in two- and three-dimensions enables new approaches to tree management and urban planning. This project will investigate novel machine learning methods to automatically detect tree species from remote sensing images.

Evolutionary machine learning and data mining

Scholarship code 649

Supervisors: Mengjie Zhang and Bing Xue

Data mining tasks arise in a wide variety of practical situations, ranging from classification to regression, clustering, and optimisation tasks. The applications range from the biomedical domain such as detecting cancers from a set of X-ray images, through the economic domain such as finding associate rules at retail sellers and predicting GDP or CPI of a nation/region, the engineering domain such as network intrusion detection and pattern matching in signal processing, to our daily life such as postal code recognition, human face detection and security control. Evolutionary machine learning techniques such as genetic programming (GP), particle swarm optimisation (PSO) and differential evolution (DE) are currently very hot topics in Artificial Intelligence (AI) and have been widely used for data mining tasks. This project aims to seek one or more students that are interested in AI and machine learning to develop and investigate new methods and algorithms using GP/PSO/DE for data mining tasks such as classification, regression, feature learning (extraction, selection and construction), unbalanced data processing, missing and incomplete data mining, clustering, text mining and natural language processing. A prospectus student is expected to take one of these areas in summer. A strong background in python/Java programming (COMP261) and in Artificial Intelligence (COMP307) is required. A good background in statistics and optimisation is desired.

Using oil wells for geothermal energy generation

Scholarship code 650

Supervisors: Ramesh Rayudu

The applicant will be involved in intensive data evaluations from different online databases. Different conceptual models will be developed from this data for different fields across the country. Our attempt is to find solutions to the intricate problems faced by the geothermal industry. The candidates are required to assist the team in comprehensive analysis of the field data and reservoir physics.  An ideal candidate will be studying a degree in Mathematics, Geology and Earth Sciences, or Engineering. Knowledge of computer language such as python, MATLAB, R, etc. with high level data visualisation skills are desirable.

Improving renewable energy usage and support for both customers and power line companies

Scholarship code 651

Supervisors: Ramesh Rayudu

This project will look at developing solutions for ways to use local renewable energy based generation amongst multiple customers and ways to improve security and resilience of the power lines. As part of the project, one student will be working on designing a community based solution to utilise majority of the generated electricity in the community. The second student will be developing a design solution for effective installation of renewable energy with a specific focus on improving the electricity network stability.

Artificial Intelligence based methods for recognising individual kākā

Scholarship code 123

Supervisor: Rachael Shaw (Biology) and Andrew Lensen (Engineering)

Identifying specific individuals is crucial for the conservation management of wild animal populations, but traditional methods require animals to be captured and tagged. This project will explore the feasibility of AI based methods for recognising specific birds in Wellington's kākā population, by testing whether an algorithm can be trained to distinguish between individual birds visiting a feeder equipped with a camera. This project will be co-supervised by Dr Andrew Lensen (Engineering) and Dr Rachael Shaw (Biology).

External projects

A web tool for visualising and performing statistical analysis on Auckland air quality data

Scholarship code 600

Supervisor: Craig Anslow and Dr Louis Boamponsem (Auckland Council)

Auckland has rich air quality data. In an age of information, visualising and discerning meaning from data is as important as its collection. The ability to derive meaningful graphics, trends, relationships from large air quality datasets depends on our access to appropriate analysis tools. Rapid data visualisation and statistical tools play an essential role in understanding and communicating results from large datasets. Effective data visualisations can support better detection, interpretation, understanding, and evaluation of information for real-time decision-making.  As data become bigger and more complex, ready-to-use visualisation techniques are crucial in discovering patterns. More importantly, interactive data visualisation allows researchers to explore data beyond what static images can offer.

Bus emissions reduction tool

Scholarship code 601

Supervisors: Craig Anslow and Chris Vallyon (NZTA)

Would you like to help improve New Zealand’s public transport?  The New Zealand government has set a target to decarbonise New Zealand public transport by 2035.  We’d like your help developing a visualisation tool that will inform scenarios to make this happen. The successful applicant will develop a bespoke interactive dashboard visualisation tool, using existing public transport data from across the country.

Emissions reduction—national inventory exercise

Scholarship code 602

Supervisors: Craig Anslow and Chris Vallyon (NZTA)

New Zealand operates more than 3,000 buses for scheduled public transport services, provided by 15 companies across 14 different Council jurisdictions. The New Zealand government has target of converting all of these buses to zero emissions vehicles by 2035. Data has been collected on the age of individual buses, and the expiration of existing service contracts.  We now need to plan the funding and staging for replacing the existing fleet. The proposed R&D would involve creating a dashboard data visualisation tool that allows future scenarios based on existing real-world data. A successful outcome might allow us to more efficiently align procurement plans with age replacement & contract expiry dates and potentially reduce the cost to the NZ taxpayer in meeting our decarbonisation targets. We could also potentially develop scenarios involving an increase in the size of the bus fleet to enable a shift from people out of single-occupancy petrol vehicles for a further decarbonisation benefit.

Design, structure and set-up of a local innovation hub

Scholarship code 603

Supervisors: Craig Anslow and Devi Alagappan (jix)

An exciting opportunity to be part of the implementation of a first of it's kind innovation hub is available for a Business/ Engineering student.  Are you sound in strategy and operational frameworks? Do you know your way around or want to work with exponential tech?  This is a transformative digital initiative to help grow an entire region. Come and be part of the change.

Transitioning to a zero carbon public transport fleet

Scholarship code 605

Supervisor: Craig Anslow and Chris Vallyon (NZTA)

Interested in improving New Zealand's Public Transport?

Injury reporting for commercial cleaners

Scholarship code 606

Supervisor: Craig Anslow and Sarah McBride (BSC)

This summer internship will be working with the Commercial Cleaning Industry body of New Zealand.  BSCNZ represents 36 large to small commercial cleaning owner/operators.  Your role will involve picking up where a fabulous group of students have left off.  Creation of an app to capture injury information on Commercial Cleaners.  This is your chance to be part of the larger project to lower the injury rates of Commercial Cleaners in NZ.  You will be working with a supportive team, combination of Victoria University senior management and the CEO of the BSCNZ Sarah McBride.  You’ll gain some amazing insights into this labour-based industry, build upon the existing work of some very talented students, and ultimately be part of this important project.

Develop a volcano rover prototype

Scholarship code 607

Supervisor: Dale Carnegie and Dr. Yannik Behr (GNS)

Build an autonomous vehicle that is capable of navigating and taking pictures of a volcanic crater like environment.

Improving hatchery algae assessments with artificial intelligence

Scholarship code 608

Supervisor: Harith Al-Sahaf and Nikki Hawes (NAI)

Aquaculture hatcheries provide a controlled environment to rear juvenile stock (e.g., mussel spat). Hatchery production and selective breeding programmes can create high-value differentiated products, but many hatchery processes are time consuming and labour-intensive, slowing the advancement of sustainable production. Microalgae are an essential food source for many hatchery reared species. Food quality can have a large impact on juvenile stock health, but assessments are often limited to estimates of algae cell concentration. Additional features that can convey food quality, such as size, morphology, pigmentation and sample contamination are often unable to be consistently assessed. This project will develop artificial intelligence (AI) technology that can rapidly analyse images of microalgae to speed up algae quality assessments, which will reduce costs, increase accuracy, and accelerate hatchery management opportunities. The aim of this project is to deliver proof-of-concept, custom-made software, incorporating state-of-the-art AI models, which automatically identifies, enumerates and measures microalgae to enable rapid quality assessment. The AI model may also include assessments of algae cell features such as cell shape and pigmentation. The challenge is to develop adaptations to object detection networks that push the limits of modern-day computer vision and AI (in particular deep learning) and apply these to a novel application.

Oyster identification using artificial intelligence

Scholarship code 609

Supervisor: Harith Al-Sahaf and Nick King (Cawthron)

Pacific oysters are key species in NZ's aquaculture industry. The visual appearance of their shells varies greatly in both colour and pattern. This project seeks to develop techniques to extract features from oyster images which will be used subsequently for individual shellfish identification to assist with breeding programme development. Image classification represents a key task in computer vision and pattern recognition. Dealing directly with the raw pixel values of images can be very difficult due to having a large search space and a single pixel can provide very little, if any, information. Aggregating a group of pixels (feature extraction) will help in reducing the search space and form a more informative and quantifiable information. Feature extraction can be performed in different scales/levels such as local (targeting a specific part of the image) and global (by considering the entire image). Furthermore, some of those extracted features can be combined together (feature construction) to build even more powerful features. However, the majority of existing feature extraction methods require domain-expert intervention to identify a good set of features that can help in categorising the instances of the different groups. Domain-experts are not always available and can be very expensive to employee. This project would use Genetic Programming (GP), a well-known and widely-investigated supervised evolutionary based algorithm, to extract shell features of oysters to explore the possibility that artificial intelligence can identify individual shellfish.

Multi-objective machine learning and applications

Scholarship code 611

Supervisor: Yi Mei and Dr Linley Jesson (Plant and Food)

In traditional machine learning, the loss function is usually a single objective that aggregates all the different aspects of consideration (e.g., accuracy for different classes, model complexity). This can hinder the performance of the machine learning methods. In this project, we aim to explore the use of evolutionary multi-objective optimisation techniques to consider the different aspects separately rather than aggregating them together. We will propose new multi-objective machine learning algorithms that can perform better than the traditional single-objective algorithms. We will test the new algorithm on some real applications faced in the Plant and Food Research Institute, such as breeding decisions, automated phenotyping and understanding biological trade-offs. To take this project, you should have a strong programming skill on Java/Python/C++, and a solid background in AI and evolutionary computation (e.g., evidenced by the grades in COMP 307/AIML 420/AIML 425/AIML 426).

Software lifecycle management tool

Scholarship code 612

Supervisor: Aaron Chen and Phil Shepherd (Harmonic)

Harmonic Analytics works with large government and company organisations who have an increasingly complex, large and diverse array of software, applications and hardware systems. These assets are typically both extremely valuable and critical to maintain for a variety of national interests. These organisations typically need to understand associated risks and prioritise investment on a quantitative basis. This project aims to develop a new Django tool that will record an inventory of software assets, risks and management plans. The tool will need to visualise how an issue with one asset affects others, and automatically create risks. Create APIs and integrate external data.

Genetic programming with transfer learning for image classification

Scholarship code 613

Supervisors: Bing Xue and Linley Jesson (Plant and Food)

Image classification is an essential component of many real-world problems. Distinguishing fishes of different species of different types, detecting tumours from brain scan images and grading fruits for retailers are just three examples. This task is very challenging due to the within-class variations and between-class similarity. It is even more difficult when the number of images for learning the classifier/model is small, where not enough information/patterns can be used to train a model that can generalise well to unseen test images. Transfer learning is an effective approach for such cases, which can use knowledge learnt from one domain (i.e. the source domain) to help learning/training in the another domain (i.e. the target domain), where only limited images are available. In aquaculture, classifying fish images of different types can be treated as transfer learning tasks. Fishes growing in different environments may share similarities but also present dissimilarities, therefore, image classification on fish images taken from different environments can be treated as different but related tasks. Using transfer learning to transfer knowledge extracted from one environment to help learning in another environment.

Network resiliency planning tool—upgrade and porting

Scholarship code 614

Supervisor: Aaron Chen and Phil Shepherd (Harmonic)

With Victoria Univesity support, Harmonic has developed a network resiliency and capacity planning tool for two telecommunications networks in NZ. Now with some international interest, Harmonic would like to enhance the tool to ensure is becomes easier and faster develop functionality for clients. We seek help to unlock the full functionality in the library and to add further functionality to the web application to support situation configuration changes.

Automated detection of Green Lipped Mussel shellfish opening

Scholarship code 615

Supervisor: Mengjie Zhang, Bing Xue and Ross Vennell (Cawthron)

It is critical for aquaculture farmers to know the condition of their crop. Currently shellfish farmers must lift their crop out of the water to inspect it. This summer project is part of a larger project aimed at developing a mussel mood monitor for the farmers. This will be based on small groups of sentinel mussels wired to measure their opening and closing as they fed on material filtered from the water. The project will develop AI techniques to automatically detect closing events within existing data from mussels in the laboratory. This project is in collaboration with the Cawthron Institute, NZ's largest independent science organisation.

Image stitching and stabilisation to improve camera AI inference for aquaculture farms

Scholarship code 616

Supervisors: Bing Xue, Ying Bi and Ross Vennell (Cawthron)

Developing new technologies to allow aquaculture farmers to monitor their farms is critical to the growth of the aquaculture industry, as many are in remote locations or far out to sea. Cameras on the farm or on vessels or drones can do this, but the number of cameras the monitor 1000’s of hectares of ocean means that developing novel AI solutions will be key to digesting large volumes of camera data. This project will test, explore, and develop “smart” processing techniques for image stitching and stabilisation, which will contribute to improving the AI inference which is already being done with the camera system in real-time. This project is in collaboration with the Cawthron Institute, NZ's largest independent science organisation.

Early detection of a marine invasive species on aquaculture farms

Scholarship code 617

Supervisor: Ying Bi, Mengjie Zhang and Ross Vennell (Cawthron)

Developing new technologies to allow aquaculture farmers to monitor their farms is critical to the growth of the aquaculture industry, as many are in remote locations or far out to sea. Cameras on the farm or on vessels or drones can do this, but the number of cameras the monitor 1000’s of hectares of ocean means that developing novel AI solutions will be key to digesting large volumes of camera data. This project will test, explore, and develop “smart” processing techniques for image stitching and stabilisation, which will contribute to improving the AI inference which is already being done with the camera system in real-time. This project is in collaboration with the Cawthron Institute, NZ's largest independent science organisation.

Mapping the predicted distribution of harmful phytoplankton after forecasted global warming

Scholarship code 618

Supervisors: Qi Chen, Ivy Liu and Ross Vennell (Cawthron)

Harmful algal blooms are proliferations of phytoplankton in the marine environment and some of these species can produce toxins that are extremely harmful to humans. Climate change has resulted in the range expansion of toxic species into new coastal regions. Using historic datasets collected over the last two decades, this project will create projected future distribution maps for high-risk species by using machine learning methods. The new knowledge generated by this project will enable predictions of toxic events and their likely risk to seafood consumers in New Zealand. This project is in collaboration with the Cawthron Institute, NZ's largest independent science organisation.

AI based signal classification for cognitive radio systems

Scholarship code 619

Supervisors: Pawel Dmochowski and Sudhir Singh (Callaghan innovation)

We have a fantastic opportunity for a summer research student to work in the Wireless Signal Processing team here at Callaghan Innovation.

Procedurally generated location-based VR game maps

Scholarship code 652

Supervisors: Craig Anslow and Jessica Manins (Beyond)

Love virtual reality? Are you an engineer keen to help Beyond research how they can procedurally generate 3D gaming environments? Come work on our research project in our cool little studio in Wellington.

Automation of photodetection linearity measurement

Scholarship Code: 653

Supervisors: James Quilty and Annette Koo (Callaghan Innovation)

The need for accurate measurements of light are important for everyday life—safe roads, productive workplaces, UV control, infrared thermometry—and for scientific research. To enable this, at the Measurement Standards Laboratory of New Zealand we are responsible for ensuring that such measurements can be made traceable to the International System of Units. For optical measurements to be meaningful, detectors must be reliable. One aspect of reliability is the linearity of their detection efficiency. In this project you will have the opportunity to apply automation, control, instrumentation, and optical design skills to the validation of photodetector linearity—see https://link.springer.com/article/10.1007/s10765-014-1743-9 for details of the system you'll be working on. In particular, you will implement stepper motor controllers, collimating optics to couple the testing system to a super continuum laser, write code to control the system and test current amplifiers.

Automation & image analysis of autocollimator

Scholarship Code: 654

Supervisors: James Quilty and Annette Koo (Callaghan Innovation)

Accurate measurements of angle are critical to quantification of light scattering. Light scattering matters when we visually assess objects for subjective qualities like freshness, quality, or realness for example, and the measurements are also used by digital artists to improve the realism of their computer-generated imagery. Instruments called goniospectrophotometers enable the characterisation of light scattered from materials using light sources, detectors, and four rotation axes. At the Measurement Standards Laboratory of New Zealand our goniospectrophotometer axes need to be aligned and calibrated for accuracy. In this project you will digitise the output of an autocollimator used to align rotation stages on the instrument. You will apply skills in design and 3D printing, image analysis, instrument control, and programming. The result will be an automated alignment process to improve the reliability of our reflectance and transmittance distribution measurements.

Contact

For further background on the scheme contact Keith Willett in the Faculty Office.

Keith profile picture

Keith Willett

Academic Coordinator, Leave, Grants and Scholars

Wellington Faculty of Science