Mechatronic engineering research internships

Supervisor: Dr Don Dansereau

Eligibility:

• Strong programming skills.
• Comfort with signals and optics maths. MECH5720 would be an asset
• An interest in optical or electronic prototyping would be an asset

Project Description:

Methane is a greenhouse gas dozens of times more potent than CO2, and most of it leaks unseen. You cannot manage what you cannot measure.

We are building a camera that sees methane. At its heart is a fibre Bragg grating tuned to the gas's spectral signature, so the detection happens optically, before any computation. That gives us one measurement at a time, and we turn it into a 2D image using computational imaging techniques. Both the reconstruction and the choice of where to look next are open problems, and machine learning has a role in each.

In collaboration with researchers at the ACFR and industry partners, this project will advance a funded prototype with a commercialisation pathway. You will work in a team spanning engineering and physics. There is scope for paper co-authorship and entry into an Honours thesis.

Depending on interest and ability you could work on:

– Optical prototyping of the imaging front end

– Single-pixel acquisition and illumination coding strategies

– Reconstruction algorithms, including learned approaches

– Active perception: learning where to look next

– Radiometric calibration and sensitivity characterisation

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

Supervisor: Dr Don Dansereau

Eligibility:

• Strong programming skills. 
• Background in analogue electronics.
• An interest in building and characterising optical and analogue electronics systems.
• Experience with PCB design or optics would be an asset

Project Description:

What does your robotic vacuum see, and who else has access to those images? In 2024, researchers showed that a robot vacuum could be hacked to stream images and audio out of someone's living room. The ABC came to us for comment, because we build a camera that never forms an image.  Rather than capturing a picture and then obscuring it, the privacy is built into the optics and the analogue electronics, before any digital signal is formed.

In collaboration with researchers at the ACFR, this project will build a hardware implementation of this approach, prototyping the analogue circuitry that accumulates light directly into a privacy-preserving representation and characterising on the bench what it can do. This line of work has produced student-authored papers and there is scope for paper co-authorship and entry into an Honours thesis.

Depending on interest and ability you could work on:

– Analogue accumulator circuit design

– Bench characterisation and mismatch measurement

– Optical and mechanical integration

– Deployment on a mobile platform  

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

Supervisor: Dr Don Dansereau

Eligibility:

• Strong Python or equivalent, and linear algebra.
• Machine learning experience would be an asset.
• Background in imaging or computer vision, AMME4710, or MECH5720 would be an asset.

Project Description:

Privacy concerns keep robots out of homes, hospitals and aged care. We build cameras that let a robot understand a room without capturing a recognisable image.

The privacy comes from how light is encoded before it is measured, so there is no image to recover. This project will design and evaluate that encoding, training it against an adversary that knows how it works and producing a privacy-utility curve that characterises its performance.

This work has been covered by Nine News, the AFR and internationally, and has produced student-authored papers. In collaboration with researchers at the ACFR, you will join a team working across optics, electronics and machine learning. There is scope for paper co-authorship and entry into an Honours thesis.

Depending on interest and ability you could work on:

– Learned optical encoding schemes

– Adversarial attacks to quantify privacy

– A nanowire mesh augmentation to the technique, in simulation

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

 

Supervisor: Dr Don Dansereau

Eligibility:

• Strong programming and mathematics skills.
• An interest in machine learning and robotic perception.
• Experience with simulation environments such as MuJoCo or PyBullet would be an asset.

Project Description:

A lunar rover drives into a dust cloud. The camera sees an obstacle where the dust has occluded the scene, the radar sees clear ground. Today's robots discard these measurements as outliers. We are building metacognition into robots, so that instead of discarding these observations, a robot makes sense of them.

This is not failure detection. We are less interested in a robot that notices it has failed than in one that revises what it believes and arrives at a better answer. Doing so means treating the disagreement as information rather than as noise.

The architecture is based on the global workspace model from neuroscience, and in collaboration with researchers at the ACFR, QUT and the University of Melbourne, this project will build and test it in 2D simulation. The work is part of a funded ARC Discovery Project, with scope for paper co-authorship and entry into an Honours thesis.

Depending on interest and ability you could work on:

– Implementing the metacognitive architecture in 2D simulation

– Belief revision and hypothesis generation

– Simulation environment and scenario design

– Evaluation of metacognitive behaviour

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

Supervisors: Dr. Don Dansereau

Eligibility:

• Strong programming skills.
• Experience with one or more of computer vision, 3D reconstruction, or ROS would be an asset.
• An interest in field deployment of a mobile robotic platform.

Project Description:

Can we keep a 3D model of a solar farm or a bridge up to date as it ages? The changes that matter are often slow and easy to miss, and capturing them reliably remains an open problem.

There is a tension between quality and frequency. A careful manual capture produces a good reconstruction but happens rarely, while a robot can visit daily but can easily miss details. This project seeks to combine the two, using infrequent high-fidelity captures to anchor frequent constrained ones, so that change can be detected and separated from distractors like seasonal variation and shifting light.

In collaboration with researchers at the ACFR and industry partners in the ARIAM Hub, this project will develop capture and reconstruction pipelines using our Scout Mini ground vehicle, 360 cameras, and Gaussian splatting. There is scope for paper co-authorship and entry into an Honours thesis.

Depending on interest and ability you could work on:

– Planning and capture on a small ground vehicle

– 360 and novel camera integration

– Gaussian splatting reconstruction

– Fusing high-fidelity and constrained captures

– Change modelling over repeat visits

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

Supervisors: Dr. Ahalya Prabhakar

Eligibility: Programming (Python, C++) and Robotics Hardware/Simulation skills required (ROS2)

Project Description:

This project will develop a motion-tracking interface that captures user demonstrations to teach robots using learning from demonstration. The project will then develop algorithmic methods for demonstration analysis for task decomposition, motion generation, and credit allocation to enable teams of robots to jointly complete a task.

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

Supervisors: Dr Ahalya Prabhakar

Eligibility: Programming (Python, C++) and Robotics Hardware/Simulation skills required (ROS2)

Project Description:

This project will develop an interactive interface that enables users to teach robots using preference-based learning. The project will develop an interface will present candidate behaviours through simulation and visualization tools, allowing users to observe how a robot performs a task and select the option that best matches their intent. These preferences will be used by learning algorithms to iteratively refine the robot’s policy, enabling the system to improve performance through human feedback. The project will provide a scalable framework for studying interactive robot learning and developing more transparent, human-centered approaches to training intelligent robotic systems.

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

Supervisors: Dr. Ahalya Prabhakar

Eligibility: Programming (Python, C++) and Robotics Hardware/Simulation skills required (ROS2)

Project Description:

This project explores how robots and interactive interfaces can support more flexible and personalized physical therapy exercises. The student will investigate how people’s reachability and movement can be measured using cameras, motion-tracking devices, or wearable sensors, and how ergodic motion planning can be used to generate therapy trajectories that encourage exploration of a patient’s reachable workspace rather than following fixed paths.

The project may involve developing an interactive tabletop, Kinect-based, or wearable-sensor interface for reaching exercises, with the longer-term goal of enabling engaging and adaptable home-based physical therapy. The project combines robotics, motion tracking, human–robot interaction, and trajectory planning, with opportunities for hands-on development and experimentation.

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

Supervisor: Dr. Ahalya Prabhakar

Eligibility:

• Programming (Python, C++)
• Machine Learning (e.g., Pytorch)
• Hardware/Simulation skills required (Computer Vision, Robot Simulators)

Project Description:

Emotive Robot Motion Planning and Control

This project explores how robots can communicate emotions and intentions through expressive movement. The student will investigate how expressive motion is created and perceived by humans, drawing on ideas from dance, theatre, gesture, and human movement analysis to identify the characteristics that make a movement appear expressive, intentional, or emotional.

These principles will then be applied to robotic motion planning and control, investigating how robots can reproduce or adapt expressive movement while respecting their physical constraints. The project may involve analysing human motion and gestures using cameras or motion-capture systems, learning representations of expressive behaviours, and developing robot trajectories that convey different emotions or intentions. The project combines robotics, human movement analysis, computer vision, motion planning, and machine learning, with opportunities for hands-on experimentation with robotic systems.

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

Supervisor: Dr. Ahalya Prabhakar

Eligibility: Programming (Python, C++) , Hardware/Simulation skills required (ROS2, Crazyflie hardware)

Project Description:

This project explores how humans can intuitively interact with and control a swarm of robots using multimodal interfaces. The student will investigate different interaction modalities—including computer vision, motion tracking, wearable sensors, and tablet-based interfaces—to allow a user to communicate high-level goals and preferences to a robotic swarm.

A key focus will be developing interfaces for ergodic swarm control, where the user can influence where and how the swarm explores an environment without needing to individually control each robot. The project may involve designing and evaluating different interaction methods, translating human inputs into swarm behaviours, and studying how effectively users can guide the swarm through gestures, body movements, or touch-based interfaces. The project combines human–robot interaction, swarm robotics, computer vision, interface design, and motion planning, with opportunities for hands-on experimentation with multi-robot systems.

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