Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests: Engineering, Computer Science, Artificial Intelligence, Machine Learning, Mathematical Modelling, and 14 moreMobile Robots, ROBOT, Imitation, Rough Terrain, Feature Extraction, Mobile Robot, Control Strategy, Terrain, Complex terrain, High Dimensionality, Range Image, Urban Search and Rescue Robotics, Urban Search and Rescue, and Autonomous mobile robot
Research Interests:
Research Interests:
Research Interests:
Research Interests:
... Andres Olave, David Wang, James Wong, Nik Von Huben, James Brooks, Tim Tam, Min Sub Kim, Alan Tay, Benjamin Leung, Albert Chang, Ricky Chen, Eric ... Eduardo Morales, Barry Drake, Achim Hoffmann, Waleed Kadous, Coral Hengst, ...
Research Interests:
Research Interests:
RoboCup inspires and motivates our research interests in cognitive robotics and machine learning, especially vision, state-estimation, locomotion, layered hybrid architectures, and high-level programming languages. The 2014 rUNSWift team... more
RoboCup inspires and motivates our research interests in cognitive robotics and machine learning, especially vision, state-estimation, locomotion, layered hybrid architectures, and high-level programming languages. The 2014 rUNSWift team comprises nal year undergradu- ate honours students, Master and PhD students, past RoboCup students and supervisors who have been involved in RoboCup for over a decade. New developments in 2014 include machine learning robot recognition, improved cascading and foveation in vision to see far-away features, re- vised locomotion and kicking, shared Kalman Filter state-estimation for localisation, a coach robot, rearchitected behaviour and skills, and chal- lenge specic innovations.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
NICTA and the New South Wales Roads and Traffic Authority (RTA) have formed a partnership to increase the effectiveness of existing traffic and transport infrastructure using smart technology. The aim is to research and develop new... more
NICTA and the New South Wales Roads and Traffic Authority (RTA) have formed a partnership to increase the effectiveness of existing traffic and transport infrastructure using smart technology. The aim is to research and develop new generation traffic control systems to tackle congestion that is estimated to currently cost Australia alone about $AU10 billion annually. This paper outlines the technical, methodological and collaborative nature of the relationship and shows how the parties complement each others experience and expertise. The research benefits from two live RTA contributed test-beds; a traffic signal controlled intersection in Sydney, and a highway roundabout in the Illawarra region. At a high level, the intended design direction is discussed, namely the development of advanced real-time traffic control systems using video sensors, traffic models and optimisation techniques for flexible signal group control.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Research Interests:
Quickly learning and recognising familiar objects seems almost automatic for humans, yet it remains a challenge for machines. This paper describes an integrated object recognition system including several novel algorithmic contributions... more
Quickly learning and recognising familiar objects seems almost automatic for humans, yet it remains a challenge for machines. This paper describes an integrated object recognition system including several novel algorithmic contributions using a SIFT feature appearance-based approach to rapidly learn incremental 3D representations of objects as aspect-graphs. A fast recognition scheme applying geometric and temporal constraints localizes and identifies the
Research Interests:
Research Interests:
Research Interests:
This paper establishes a framework that hierarchically integrates symbolic and sub-symbolic representations in an architecture for cognitive robotics. It is formalised abstractly as nodes in a hierarchy, with each node a sub-task that... more
This paper establishes a framework that hierarchically integrates symbolic and sub-symbolic representations in an architecture for cognitive robotics. It is formalised abstractly as nodes in a hierarchy, with each node a sub-task that maintains its own belief-state and generates behaviour. An instantiation is developed for a real robot building towers of blocks, subject to human interference; this hierarchy uses a node with a concurrent multi-tasking teleo-reactive program, a node embedding a physics simulator to provide spatial knowledge, and nodes for sensor processing and robot control.
Research Interests:
A mobile robot must know where it is to act appropriately. An algorithm that allows a robot to accurately localise itself locally using a vision sensor and a map of its environment is described in this paper. The basic idea of this... more
A mobile robot must know where it is to act appropriately. An algorithm that allows a robot to accurately localise itself locally using a vision sensor and a map of its environment is described in this paper. The basic idea of this algorithm, called NightOwl, is to match the projected camera image with a map of the environment in a local area in order to find the most likely position and orientation of the camera platform.
This paper establishes a framework that hierarchically integrates symbolic and sub-symbolic representations in an architecture for cognitive robotics. It is formalised abstractly as nodes in a hierarchy, with each node a sub-task that... more
This paper establishes a framework that hierarchically integrates symbolic and sub-symbolic representations in an architecture for cognitive robotics. It is formalised abstractly as nodes in a hierarchy, with each node a sub-task that maintains its own belief-state and generates behaviour. An instantiation is developed for a real robot building towers of blocks, subject to human interference; this hierarchy uses a node with a concurrent multitasking teleo-reactive program, a node embedding a physics simulator to provide spatial knowledge, and nodes for sensor processing and robot control.
This report describes the development and implementation of the locomotion for the Nao H25 V4 robot as used by team rUNSWift from the University of New South Wales, Australia, for the 2014 RoboCup SPL competition. We refer to the... more
This report describes the development and implementation of the locomotion for the Nao H25 V4 robot as used by team rUNSWift from the University of New South Wales, Australia, for the 2014 RoboCup SPL competition. We refer to the omnidirectional locomotion motion as a walk. The main purpose of the report is to document the 2014 walk for the UNSW 2014 SPL code release. School of Computer Science & Engineering University of New South Wales Sydney 2052, Australia
ABSTRACT Without a model the application of reinforcement learning to control a dynamic system can be hampered by several shortcomings. The number of trials needed to learn a good policy can be costly and time consuming for robotic... more
ABSTRACT Without a model the application of reinforcement learning to control a dynamic system can be hampered by several shortcomings. The number of trials needed to learn a good policy can be costly and time consuming for robotic applications where data is gathered in real-time. In this paper we describe a variable resolution model-based reinforcement learning approach that distributes sample points in the state-space in proportion to the effect of actions. In this way the base learner economises on storage to approximate an effective model. Our approach is conducive to including background knowledge to speed up learning. We show how different types of background knowledge can used to speed up learning in this setting. In particular, we show good performance for a weak type of background knowledge by initially overgeneralising local experience.
Research Interests:
Research Interests:
Research Interests:
Research Interests:
RoboCup continues to inspire and motivate our research in- terests in cognitive robotics and machine learning, particularly layered hybrid architectures, abstraction, and high-level programming languages. The newly formed 2010 rUNSWift... more
RoboCup continues to inspire and motivate our research in- terests in cognitive robotics and machine learning, particularly layered hybrid architectures, abstraction, and high-level programming languages. The newly formed 2010 rUNSWift team includes mainly nal year un- dergraduate students under the supervision of leaders who have been involved in RoboCup for many years. In 2010 the team revamped the entire code-base
Research Interests:
Multi-agent robotic competitions such as RoboCup provide the motivation for a developmental research agenda–one that focuses on the evolution of complete working systems and their cognitive architectures. In this paper, we describe the... more
Multi-agent robotic competitions such as RoboCup provide the motivation for a developmental research agenda–one that focuses on the evolution of complete working systems and their cognitive architectures. In this paper, we describe the components and integration of one such system–the 2010 RoboCup Standard Platform League entry rUNSWift. The realtime control architecture employed consists of a hierarchy of modules that implement functions of perception, world-modelling and behaviour generation. The ...
Competitive bipedal soccer playing robots need to move fast and react quickly to changes in direction while staying upright. This paper describes the application of reinforcement learning to stabilise a flat-footed humanoid robot. An... more
Competitive bipedal soccer playing robots need to move fast and react quickly to changes in direction while staying upright. This paper describes the application of reinforcement learning to stabilise a flat-footed humanoid robot. An optimal control policy is learned using a physics simulator. The learned policy is supported theoretically and interpreted on a real robot as a linearised continuous control function. The paper also describes other components, including foot-step coordination, of bipedal locomotion integrated to achieve reactive omni-directional locomotion for Nao robots used in the RoboCup Standard Platform League.
An open problem in reinforcement learning is discovering hierarchical structure. HEXQ, an algorithm which automatically attempts to decompose and solve a model-free factored MDP hierarchically is described. By searching for aliased Markov... more
An open problem in reinforcement learning is discovering hierarchical structure. HEXQ, an algorithm which automatically attempts to decompose and solve a model-free factored MDP hierarchically is described. By searching for aliased Markov sub-space regions based on the state variables the algorithm uses temporal and state abstraction to construct a hierarchy of interlinked smaller MDPs.