������ýapp

In partnership with , we’re developing a network of mobile robots controlled over a 5G link in near real time. This project is investigating:

• typical characteristics of 5G (extremely low latency, high reliability and fast throughput) 

• the industrial applications of real-time control and automation enabled by 5G

The mobile robots we’re working with can accomplish tasks such as parts delivery, product inspection and collaboration with human operators or other robots during the manufacturing process. 

We’re developing mobile robotic systems to operate in unstructured environments, specifically orchard, agricultural and forestry environments.

The system features onboard sensors (including an inertial measurement unit and computer vision), superior mobility, agility and intelligence, and can be automated to undertake a variety of operations, such as assisting orchard workers, transporting, spraying, weeding, harvesting, data collection and environmental management. 

Our research covers these topics:

• novel locomotion and mobility of ground vehicle systems

• modelling and disturbance-free proximity control of miniature aerial vehicles for environment and infrastructure monitoring and management

• deep learning pattern recognition for robotic guidance in natural environments

• error modelling and quantification of imaging systems to improve localisation

• multi-sensor data fusion for accurate localisation and positioning

• multi-agent cooperative control for agricultural and post-farm gate logistics

In partnership with , we’re exploring the adaptation of robotic vision techniques to establish a data-driven pipeline for enhancing wearable assistive technologies such as retinal implants. 

Retinal implants, also known as bionic eyes, are the leading treatment for retinal diseases such as Retinitis Pigmentosa and Age-related Macular Degeneration. This implantable micro-electrode array can restore a sense of vision by electrically stimulating surviving retinal cells. The primary input to this process is video frames acquired from a head-mounted camera.

However, due to the highly restricted resolution of state-of-the-art implants (as low as 40 pixels), a significant down-sampling must be performed on the image frames captured. This has led us to consider computer vision and image enhancement algorithms to detect, extract and augment the signal passed to the implant in order to emphasise key features in the environment that are crucial to safe and independent visual guidance for the implant recipient. ‌

In our research, we have explored the adaptation of robotic vision techniques, such as:

• ground plane segmentation

• human vision inspired vision algorithms for motion processing 

• state-of-the-art deep learning and deep reinforcement learning

We’ve established a data-driven pipeline for both learning and enhancing visual features of importance to high priority visuo-motor tasks, such as navigation and table-top object grasping. Through patient testing and simulated testing with virtual reality, we aim to develop and thoroughly evaluate vision processing solutions that improve functional outcomes, enhance independence and provide useful vision for patients implanted with these devices.

An example of vision processing and wearable assistive technologies in use.

In partnership with and , we’re developing several multi-agent cloud robotic systems that can be used in environments, such as school classrooms or to provide first aid assistance for elderly people.

The introduction of cloud robotics has merged the two ever-progressing domains of robotics and cloud computing. The added feature of cloud implies less dependence on human input and more support from ubiquitous resources. Industry 4.0 is envisioned to be a key area for infusion of these robotic technologies, especially in automating applications such as sensing, actuating and monitoring via insurgence of cloud computing and wireless sensors. Particularly, smart factories located in remote locations with challenges to health, safety and environment are our motivation for robotic inspection, maintenance and repair.

Cloud-aided robots can complement the sensors with action-oriented task, such as inspection, fault diagnosis and sensor testing. The tasks associated with these robotic applications are usually interdependent, latency sensitive and compute intensive. In addition, the resources are heterogeneous in terms of processing capability and energy consumption. These smart factory applications require robots to continuously update intensive data in order to execute tasks in a coordinated manner. Therefore, real-time requirements need to be fulfilled by tackling resource constraints.

Given the context, the specific aim of our research team is to design real-time resource allocation schemes to improve resource sharing among robots and optimise task offloading to the cloud for multi-agent cloud robotic systems. We’ve already implemented evolutionary approaches to make offloading decision for a single robot factory maintenance application in a sample oil factory workspace and optimal resource allocation for the tasks of emergency management service in a smart factory.

The research group has also worked on several proof of concepts, including the development of a cloud-aided robot hardware, face recognition, voice-based control, environment mapping and path planning via cloud-supported robot. We are also supporting elderly people with first aid assistance using Turtlebot 2.0, as well as integrating robot teaching assistance in a school classroom. 

An example of a multi-agent cloud robotic system.

Collaborative robots (cobots) are designed to work safely in close proximity to humans. The introduction of low-cost cobots has enabled new robot applications in multiple new domains including healthcare. However, programming a robot still requires expert knowledge and the real revolution in cobot use will happen first when teaching a robot to perform a new task is simplified.

Funded through the Australian Research Council Training Centre for Collaborative Robotics in Advanced Manufacturing, our research is focused on making cobots learn tasks from human demonstration.

Funded through the Australian Research Council Discovery Projects scheme, our research is focused on:

• understanding the dynamic behaviours of robotic control systems in networked environments 

• developing a new theory for the analysis and design of networked robotic control subject to network-induced constraints

Robotic control is used in many industry applications to improve efficiency, productivity and accuracy. The development of digital technology, especially Industry 4.0, brings in more networked control systems that create new design challenges due to network-induced constraints such as delay, data packet dropouts and cyber attack.

We’re investigating smart algorithms for robotic and mechatronic systems subjected to nonlinearities, disturbances and uncertainties. The research includes modelling and pattern recognition, control and optimisation, sensoring and data processing.

To date, our developed algorithms have been used in electric vehicle steer by wire systems, anti-lock brake systems, flexible robotic joints, industrial drives, unmanned autonomous vehicles and more.

A collaborative robot (or cobot) is a robot which has been specifically designed to work closely with humans without the need for additional safety equipment or fencing. To achieve this, the cobot is typically equipped with a variety of force and position sensors that detect the presence of humans and objects in the workspace.

This suite of sensors provides an opportunity for the robot system to be re-designed to evolve from a “blind” tool to an intelligent machine that can develop a “sense” of its environment and self-monitor the outcomes or products it creates.

This project will develop methods by which the cobot will self-evaluate its performance of tasks and outcomes in real time using information gathered from how it perceives motions and actions. Ultimately, it would be able to evaluate its performance in much the same way that a skilled craftsperson (e.g. welder, woodworker, etc.) knows by experience that they have produced a good quality product.

As humans adapt and familiarise with their cobot teammates, emergent work patterns and  behaviours are likely to evolve and be subject to change. Similarly, augmentations to the cobots workflow patterns may emerge from AI-driven adaptations to the human operator’s own changing patterns and behaviours, or simply due to upgrades to the system or changes to the underlying working conditions.

This presents the challenge of both balancing the need to maintain and monitor potentially stringent compliance requirements, while also allowing sufficient scope for the human-robot collaboration to adapt to the needs of the task and to establish efficient collaborative workflows. This project will explore strategies to account for such needs and keep process outputs within the bounds of compliance.

In partnership with the , we’re exploring the use of teleoperation for medical diagnosis, specifically for heart ultrasounds in remote areas. Our project targets all aspects of robot-assisted remote ultrasound examinations.

Although medical advances are extending the life expectancy of the Australian population, access to health centres, hospitals and medical specialists can vary depending on where people live. Those in major cities or urban areas typically have excellent access to the relevant facilities and specialists; those in remote and rural areas are typically less fortunate. 

One potential solution is the teleoperation of medical procedures. This technology is particularly important for a large, sparsely populated country like Australia. Our project utilises teleoperation for medical diagnosis, specifically for ultrasounds of the heart (an echocardiogram).

These use high-frequency sound waves to capture images of a heart, enabling the cardiographer to study the heart’s functions, such as the pumping strength, the working of the valves and to check if there are any tumours or infectious growth. An echocardiogram is a very important tool to diagnose heart conditions such as heart murmurs, enlarged heart and valve problems. 

In partnership with  (RCH) and , we've been developing a socially assistive humanoid robot as a therapeutic aid for paediatric rehabilitation. The aims are to motivate young patients, increase exercise adherence and deliver rehabilitation care efficiently.

After three and half years of development and testing, the robot has engaged with over 50 patients, predominantly those with cerebral palsy, and is undergoing evaluation with therapists and parents within RCH’s rehabilitation clinic. 

Adapting and deploying a general-purpose social robot as a therapeutic aid for diverse young patients with varying needs in a busy hospital setting comes with significant challenges. This is not typically addressed in traditional human-robot interaction or robotics research generally.

This project is thus exploring how state-of-the-art artificial intelligence and computer vision can be deployed long term in real-world clinical settings to augment therapy delivery and improve patient outcomes.

We’re working with groups from the  to find out what robots can contribute to the quality of life for older adults and adults living with dementia. Our team of interaction designers, psychologists and engineers are interested in people’s attitudes to, and relationships with, robots.

We’re exploring which characteristics increase acceptance and adoption in everyday life and therefore create benefits for the user, as well as researching and discussing ethical considerations and societal issues surrounding the role of robots.

Traditionally, a strong focus in robot development concerns the functional aspects of robots. The design aesthetics of robots and their outputs (speech, gestures and emotions) are often overlooked or neglected.

We’re interested in speech interactions, the visual appearance of robots and their ability to convey emotions, so we’re exploring the impact of improved speech interfaces and communication patterns of robots with users and how these support older adults to live independently at home for longer.