Reinforcement Learning
Our research explores reinforcement learning (RL) for developing autonomous systems that can learn and adapt in complex, changing environments. We develop new RL algorithms and methods with a focus on robust, adaptable and generalisable behaviour, supported by our open-source CARES Reinforcement Learning framework for developing, evaluating and benchmarking RL approaches across simulated and real-world robotic applications.
We work across all robotics domains: land, air, sea, and space with various research and industry collaborators.
See our Github for full Details:
https://github.com/UoA-CARES/cares_reinforcement_learning
Field Robotics
Our research in agricultural robotics develops autonomous systems for challenging and unstructured field environments. We have worked across kiwifruit, apple, grapevine and blueberry crops, developing robotic systems for tasks including harvesting, pollination and pruning. Our work brings together robotic perception, manipulation, navigation and intelligent control to address practical challenges in horticultural automation.
Underwater Robotics
Our research in underwater robotics develops autonomous robotic systems for operation in challenging marine environments. We explore underwater perception, navigation and control, with applications including autonomous ship hull inspection, the detection and management of marine biofouling, and robust navigation where conventional positioning systems are unavailable.
Dexterous Manipulation
Our research in dexterous robotic manipulation explores how robots can perceive, grasp and interact with objects in complex and unstructured environments. We develop methods combining vision, tactile and touch sensing with learning and intelligent control to enable adaptive, precise and robust manipulation across a diverse range of objects and tasks.
See our Github for full details:
https://github.com/UoA-CARES/gripper_gym
Autonomous Racing (F1Tenth/FSAE)
Our research in autonomous racing explores high-performance perception, planning and control at the limits of vehicle handling. Using platforms including F1TENTH and Formula SAE, we develop and evaluate learning-based and autonomous driving methods for high-speed navigation, trajectory planning, vehicle control and decision-making in dynamic racing environments.
See our Github for full details:
https://github.com/UoA-CARES/autonomous_f1tenth
Autonomous Drones
Our research in aerial robotics develops autonomous drones capable of navigating and operating in complex and dynamic environments. We explore perception, localisation, navigation, planning and intelligent control, including learning-based approaches for robust flight in challenging environments where conventional sensing and positioning may be limited.
See our Github for full details:
https://github.com/UoA-CARES/drone_gym
Space Systems
Our research in space robotics explores autonomous guidance, navigation and control for spacecraft and robotic systems operating in challenging space environments. We investigate intelligent and learning-based control approaches for robust autonomous operation, with applications including spacecraft manoeuvring, attitude control and proximity operations.