Electrical and computer engineering

Supervisor: A/Prof. Steve Shu

Eligibility: Familiarity with deep learning frameworks, Signal processing and analysis and mathematical modelling

Project Description:

Deep learning-based diffractive imaging technology, such as PtychoNN, offers phase recovery hundreds of times faster than traditional algorithms (like Ptycholib). However, the latest synchrotron instruments, with their significantly increased data rates, demand approximately PFLOPs of computation for phase retrieval, posing a challenge for real-time imaging.

To meet this challenge without compromising performance, this project aims to optimize lightweight networks for inverse imaging via pruning, quantization, and knowledge distillation, thereby accelerating phase recovery tasks.

Reference:

[1] Pan, Xinyu, et al. "An efficient ptychography reconstruction strategy through fine-tuning of large pre-trained deep learning model." Iscience 26.12 (2023).

Requirement to be on campus: No

Supervisors: A/ Prof. Steve Shu

Eligibility: fundamentals of optoelectronics; Proficient in programming languages (Matlab or Python)

Project Description:

Ptychography was originally proposed to solve the phase problem in electron crystallography. In the past decade, it has evolved into an enabling imaging technique for both fundamental and applied sciences. In this project, in order to solve the inherent tradeoff between optical system resolution and field of view, we will build a new imaging system, lensless coded ptychography approach.

Reference:

[1] Liming Yang, et. al., “Lensless polarimetric coded ptychography for high-resolution, high-throughput gigapxiel birefringence imaging on a chip,” Photonics Research, 11(12), 2242-2255 (2023)

Requirement to be on campus: Yes *dependent on government’s health advice

Supervisor: Dr Cuo (Charlie) Zhang

Eligibility:  Solid skills of Matlab/Python programming and data analysis

Project Description:
Large language models, such as generative pre-trained transformer (GPT), have been widely used in our daily life, and it is a great opportunity to apply a GPT agent for decision making on power system operation and control.

This project will develop a model of linguistic stipulations, which contains context, question and answer, as an agent. Then, this project will design a method of presenting the power system operating conditions via this agent and interacting with GPT to improve the performance of decisions.

This GPT assisted method is designed for targeted power system problems including volt/var control in an unbalanced distribution network and optimal power flow in a microgrid.

Requirement to be on campus: No

Supervisor: A/Prof. Dong Yuan

Eligibility: Good programming skill. Familiar with UE5 and C++ are preferrable.

Project Description:

Wireless technology as the backbone of mobile applications has become essential in our daily lives. However, signal fluctuations caused by the unpredictable nature of factors, such as moving crowds and improvised events in wireless and mobile applications amplify the complexity of their comprehensive evaluation across diverse real-world environments.

To address these challenges, we developed a digital twin system with learning-based calibration and ray-tracing based real-time wireless propagation simulation capabilities for WMAs. The system can learn the latent state of the environment from real data for high-precision calibration and accurately simulate wireless systems in realistic 3D environment.

The candidate will work on the applications of this system, e.g., autonomous driving and smart manufacturing, which will use measurement of wireless signals for training AI models.

Requirement to be on campus: No

Supervisor: A/Prof. Dong Yuan

Eligibility: Familiar with PyTorch and Nvidia CUDA; programming with Python and C++

Project Description:

Recent advances in deep neural networks (DNNs) have substantially improve the accuracy and speed of video analytics. The maturity of cloud computing, equipped with powerful hardware like GPU, becomes a typical choice for such kind of computation intensive DNN tasks. One obstacle, however, is the large amount of data volume of video streams.

For example, a self-driving car can generate up to 750 megabytes of sensed data per second, but the average uplink rate of 4G, fastest existing solution, is only 5.85 Mbps. In order to avoid the effects of network delay and put the computing at the proximity of data sources, edge computing emerges. Nevertheless, edge computer itself is limited by its computing capacity and energy constraints, which cannot fully replace cloud computing.

This project will investigate the efficient parallel algorithms for DNN inference tasks on the edge server that equipped with GPUs.

Requirement to be on campus: No

Supervisor: Dr Wibowo Hardjawana

Eligibility: Strong background in wireless communication and deep learning, equivalent to the one covered in ELEC5508 Wireless Engineering

Project Description:

Each generation of cellular communication systems is marked by a defining disruptive technology of its time, such as orthogonal frequency division multiplexing (OFDM) for 4G or Massive multiple-input multiple-output (MIMO) for 5G. Since artificial intelligence (AI) is the defining technology of our time, it is natural to ask what role it could play for 6G.

The project aims to demonstrate a 6G vision of a new air interface and AI replacing communication blocks to enable optimised end-to-end communication schemes for any hardware, radio environment, and application. The specific tasks of the project are to modify and/or add existing wireless communication reference design to facilitate end-to-end learning by including AI in the wireless air interface. Students will need to compare its performance with a standard communication system.

Requirement to be on campus: Preferrable on Campus

Supervisor: Prof. Philip Leong

Eligibility: Student must have completed ELEC3608 (or an equivalent course in computer architecture) and have strong FPGA design skills.

Project Description:

The computational complexity of large language models (LLMs) is a limiting factor for their widespread adoption. Recent research has enabled quantisation of the weights to ternary (the weights are in the set {-1,0,+1}) values without loss of accuracy https://arxiv.org/abs/2402.17764. This result paves the way for improved latency, memory, throughput, and energy consumption and the same paper says, “we envision and call for actions to design new hardware and systems specifically optimized for 1-bit LLMs”.

Examples of previous ternary neural network accelerators include xTern https://arxiv.org/abs/2405.19065, a RISC-V extension and our own ternary neural network tool http://phwl.org/assets/papers/ternary_trets19.pdf

In this research, we will develop an improved ternary neural network FPGA-based accelerator for small LLMs and quantify the advantages.

Requirement to be on campus: No

Supervisors: A/Prof. Weidong Xiao

Eligibility:

• Strong in power electronics and control engineering
• Hand-on experience for prototyping is a plus.

Project Description:

Collaborated with a high-tech company, CGD, the project aims to develop a prototype of a high-efficiency converter using wide-band-gap power semiconductors.

One or two students are required to work with one industry engineer for the development. The facility of the power electronic research lab is available for the development. The student(s) shall gain real-world industry experience, which is good for the future career.

Requirement to be on campus: Yes, work in research lab *as per government’s health advice.

Supervisors: A/Prof. Zihuai Lin, Dr Audrey Wang, Prof. Branka Vucetic

Eligibility: up to 2 students are required for this project. The students participating in this project should have good knowledge on hardware development. Programming skills are essential. The students with average marks above 75 are preferred.

Project Description:

This project will develop novel solutions for IoT based pressure injury monitoring to enable smart hospitals, smart healthcare and provide in-patients with good treatment experience. IoT is a key enabler for our future smart hospital and healthcare to effectively overcome the major problems, such as nursing staff shortness, impractical physical care environments and difficulties in identifying patents’ needs, etc. 

In order to enable a fast uptake of the IoT pressure injury monitoring systems, key issues, such as data collection and sensing, pressure injury prediction modelling,  and hospital validation, should be addressed. These problems are the major technological obstacles which are preventing industrial partners from further expanding their business in the hospital IoT area.

We will develop a smart mat for pressure injury monitoring. The proposed smart mat consisting of a pressure sensor array and a moisture sensor array will be tested and validated in the Westmead hospital. The collected data will be transmitted to the cloud via WiFi or other wireless networks for data processing and analysis. The developed deep learning algorithms for smart mat will be used to obtain the weight and moisture distribution of the inpatient’s body as well as the respiratory rate.

Requirement to be on campus: Yes *as per government’s health advice.

Supervisors: A/Prof. Zihuai Lin and Prof. Branka Vucetic.

Eligibility: up to 2 students are required for this project. The students participating in this project should have good knowledge on wireless cellular networks, communication theory and signal processing. Matlab, C++ and Python programming skills are essential. The students with average marks above 75 are preferred.

Project Description:

The ambient backscatter communications (AmBC) technology used in IoT systems is a practical application of energy harvesting from ambient radio frequency (RF) signals.

In a typical AmBC-empowered IoT system, a passive or semi-passive IoT device harvests RF energy from the ambient RF source (e.g., a cellular network base station (BS) or a Wi-Fi router) to power its circuits, and then the IoT device modulates its information bits onto a sinusoidal signal by intentionally changing its amplitude and/or phase. In this project, we will build an AmBC-empowered battery-free IoT device prototype.

Requirement to be on campus: Yes *as per government’s health advice.

Supervisors: A/Prof. Zihuai Lin and Prof. Branka Vucetic

Eligibility: up to 2 students are required for this project. The students participating in this project should have good knowledge on wireless cellular networks, communication theory and signal processing. Matlab, C++ and Python programming skills are essential. The students with average marks above 75 are preferred.

Project Description:

With explosive growth of wireless networks and the Internet, network resource utilization becomes one of the critical design issues. RIS has emerged as a promising tool for the design of next generation communication networks. There is a need to develop systematic design of RIS THz array techniques for real wireless energy efficient networks.

In this project, we plan to systematically design RIS channel estimation, beamforming mechanisms for cellular networks and related signal processing algorithms, as well as demonstrate the feasibility and benefits of integrating the developed RIS schemes into practical systems. The research outcomes are likely to result in significant improvements in network performance.

Requirement to be on campus: Yes *as per government’s health advice.

Supervisor: A/Prof. Zihuai Lin

Eligibility: up to 2 students are required for this project. The students participating in this project should have good knowledge of hardware design, Electronic Engineering, and signal processing. The students with average marks above 75 are preferred.

Project Description:

Electroacupuncture is a method of preventing and treating diseases by combining needle and electrical stimulation, after inserting needles into acupoints to obtain qi, passing a small amount of current close to the human body's bioelectricity through the needles.

After the needle is inserted into the acupoint to obtain qi, a small electric current wave of human bioelectricity is passed through the needle (induced), which is divided into: 1. Continuous wave 2. Intermittent wave. It is a therapy to stimulate the acupoint and treat diseases. In this project, we plan to develop a wireless powered acupuncture system, design intelligence electronics acupuncture needles.

Requirement to be on campus: Yes *as per government’s health advice.

Supervisors: Dr Darshana Jayasinghe, Prof. Sri Parameswaran

Eligibility: WAM>75, Verilog or VHDL knowledge

Project Description:

JitSCA (Jitter-based Side-Channel Analysis) is an attack used by adversaries to break into encrypted data by measuring very tiny timing changes in computer operations. These tiny changes, called jitter, can reveal secrets even if the encryption algorithm is strong. Normally, these tiny timing variations are ignored, but JitSCA uses them to attack cryptographic systems that protect sensitive information like credit card numbers and personal data.

In this short research, we are working on improving JitSCA attacks by using different hardware designs to measure jitter. We've created three new hardware designs which we think will help us perform JitSCA attacks more effectively. We will test these designs on an FPGA development board (e.g., a Digilent ZedBoard) to investigate the best hardware design to detect the timing changes most accurately. We'll first replicate previous work to set a baseline and then compare our new designs to find the best one.

Requirement to be on campus: Yes, *as per government’s health advice.

Supervisors: Dr Darshana Jayasinghe, Prof. Sri Parameswaran

Eligibility: WAM > 75, Verilog or VHDL knowledge

Project Description:

In Remote Power Analysis (RPA) attacks, attackers use special hardware designs to measure how much power circuits use. By studying these power patterns, they can figure out the secret keys used in the encryption.

Currently, attackers measure power consumption and then analyse it offline (meaning not in real-time) to reveal the secret keys. This method is effective even on cloud-based FPGAs.

In our research, we aim to make this process more direct by doing the attack directly on the FPGA itself, rather than analysing the power consumption data transmitted to another location for processing. We will use the Pearson Correlation Coefficient (PCC) algorithm, a statistical method to process power consumption data, to perform the RPA attacks on the FPGA. We already have the code needed to measure power usage, and your task will be to implement the PCC algorithm on the FPGA to conduct the attack.

Requirement to be on campus: Yes *as per government’s health advice.

Supervisor: Prof. Jian Guo Zhu

Eligibility:

• Bachelor's degree in electrical engineering
• Outstanding transcript (WAM>75)
• Good knowledge of electromagnetics
• Skillful with MATLAB/Simulink or other programming languages
• Research experience with a publication record in electrical engineering (preferred)

Project Description:

System-level design optimization of electromagnetic devices requires fast solutions to multiphysics field problems. This is currently done numerically using methods like finite element analysis, which takes an excessively long time.

This project aims to develop an AI-based method to solve partial differential equations, laying a solid foundation for multiphysics/multidisciplinary engineering field problem analysis.

Requirement to be on campus: No

Supervisor:  Prof. Xiaoke Yi, A/Prof. Luping Zhou, A/Prof. Liwei Li

Eligibility: Year 3/4/5 or Master students, Electrical engineering, Mechatronics, Computer engineering,

Project Description:

The state-of-the-art sensing technology is rapidly growing and will play a critical role in the near future. For instance, smart phones, which play a significant role in our daily lives, have a fingerprint identity sensor that makes it easy for us to access the device, and they also use an ambient brightness sensor to adjust the display brightness, etc.

The project is to deliver superior, advanced sensing platforms assisted by machine/deep learning to address the important challenges across a diverse range of applications in various fields, particularly in lab-on-chips, Internet of Things, broadband communications and biomedical applications. The internship project focuses on electrical circuit design and data processing as well as machine learning and software programming. The aim is to realise ultra-sensitive, high resolution and extreme-range sensing.

The intern will work closely with a research team that includes PhD students and postdoctoral research associates. Innovative signal processing and design in both hardware and software will be carried out during the project.

Requirement to be on campus: Yes, *dependent on government’s health advice.

Supervisors: Prof. Xiaoke Yi, A/Prof. Liwei Li

Eligibility: Year 3/4/5 or Master students, Electrical engineering, mechatronics, computer engineering, software engineering or computer science

Project Description:

Electro-optic modulators encode electrical signals onto an optical carrier. They are essential for the operation of global communication systems and data centres for artificial intelligence, broadband networks, and high-performance computing.

The project focuses on the development of an ultra-wideband electro-optic modulator. Addressing the challenges associated with achieving a large modulation bandwidth entails reducing microwave attenuation and realizing velocity and impedance matching. It is also essential to optimize the modulation electrode and optical waveguide jointly. The project will advance signal modulation techniques, paving the way for enhanced optical communication and modulation capabilities.

Requirement to be on campus: Yes, *dependent on government’s health advice.

Supervisor: Prof. Xiaoke Yi

Eligibility: Year 3/4/5 or Master students, Electrical engineering, mechatronics, computer engineering, software engineering or computer science

Project Description:

In the pursuit of enhancing imaging technology to new heights, the project aims at developing a new camera that embodies compactness, an expansive field of view, and low power consumption.

Leveraging the latest advancements in photonic technology, the primary focus lies in the design and testing of the camera for high performance optical imaging.  The intern will work with a collaborative team, comprising PhD students and research associates in this project.

Requirement to be on campus: Yes, *dependent on government’s health advice.