Chris Gaskett

Patent: 視線座標変換方法、それのプログラムおよびそれを記録した記録媒体

Japanese patent application 2002-38390, granted 2003-241892. Rights owned by Toyota Central Research and Development Laboratory.

Is Optimal Foraging a Realistic Expectation in Orb-web Spiders?

Ecological Entomology, Vol 34, 2009, pp 527-534

1. Explanations for web relocation invoking optimal foraging require reliable differentiation between individual sites and overall habitat quality. We characterised natural conditions of resource variability over 20 days in artificial webs of the orb-web spider Gasteracantha fornicata to examine this requirement. 2. Variability in catch success was high. Day-to-day catch success in 90% (18/20) catch sites fitted negative binomial distributions, whereas 10% fitted Poisson distributions. Considered across trap sites (overall habitat), variance in catch success increased proportionally faster than the mean (i.e. Taylor's Power Law, variance = 0.54 mean 1.764). 3. We compared the confidence intervals for the expected cumulative catch in randomly drawn sequential samples from a frequency distribution representing the overall habitat (based on the parameters for Taylor's power law) and the frequency distribution of expected cumulative catch within each individual catch site [via randomisation based on the mean and negative binomial exponent (k)]. 4. In all cases and across all sample sizes, median values for the power to differentiate habitat and catch sites never exceeded 0.2, suggesting that principles involved in optimal foraging, if operating, must be accompanied by a very high degree of uncertainty. 5. Under conditions of high resource variability, many days must be spent in a single catch site if movement decisions are based on an ability to differentiate current catch site from overall habitat. Empirical evidence suggests this is never met. This may explain why proximal mechanisms that illicit quickly resolved behavioural responses have been more successful in describing web relocation patterns than those associated with optimal foraging.

Creativity in the Cane Fields: Motivating and Engaging IT Students Through Games

Proceedings of the 2nd Annual Microsoft Academic Days Conference on Game Development, (aboard Disney Wonder Cruise Ship, 2007)

Foveated Vision Systems with Two Cameras per Eye

Proceedings of the IEEE/RSJ IEEE Int. Conf. Robotics and Automation (ICRA 2006), (Orlando, Florida, May 2006)

We present an exhaustive analysis of the relationship between the positions of the observed point in the foveal and peripheral view with respect to the intrinsic and extrinsic parameters of both cameras and 3-D point position.

pdf document

Hand-Eye Coordination through Endpoint Closed-Loop and Learned Endpoint Open-Loop Visual Servo Control

The International Journal of Humanoid Robotics, Vol 2, No 2, 2005 pp. 203-224

We propose a hand-eye coordination system for a humanoid robot that supports bimanual reaching. The system combines endpoint closed-loop and open-loop visual servo control. The closed-loop component moves the eyes, head, arms, and torso, based on the position of the target and the robot's hands, as seen by the robot's head-mounted cameras. The open-loop component uses a motor-motor mapping that is learnt online to support movement when visual cues are not available.

Support Vector Machines and Gabor Kernels for Object Recognition on a Humanoid With Active Foveated Vision

Proceedings of the IEEE/RSJ Int. Conf. Intelligent Robots and Systems (IROS 2004), (Sendai, Japan, September 2004)

pdf document

Online Learning of a Motor Map for Humanoid Robot Reaching

Proceedings of the 2nd International Conference on Computational Intelligence, Robotics and Autonomous Systems (CIRAS 2003), (Singapore, December 2003)

pdf document

movie (6.1MB)

Reinforcement Learning Under Circumstances Beyond its Control

Proceedings of the International Conference on Computational Intelligence for Modelling Control and Automation (CIMCA2003), (Vienna, Austria, February 2003)

This paper describes an RL algorithm that uses an idea from decision theory.

pdf document

Q-Learning for Robot Control

PhD thesis from the Robotic Systems Laboratory, part of the Research School of Information Sciences and Engineering at the Australian National University.

As well as more detailed versions of the material below, the thesis includes a partial solution to the rising value problem identified by Thrun and Schwartz (1993). It also contains a longer, more up to date survey of other continuous state and action reinforcement learning algorithms than the papers.

pdf document

Learning Implicit Models during Target Pursuit

Proceedings of the IEEE International Conference on Robotics and Automation (ICRA2003), (Taiwan, May 2003).

This paper demonstrates control of one joint of an active head and shows that the neural network has learnt a model of movement under gravity.

pdf document

short movie (3.3MB)

Reinforcement Learning for a Vision Based Mobile Robot

Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2000) © IEEE (Takamatsu, Japan, October 2000).

This paper applied the algorithm to wandering and servoing for a mobile robot. It also demonstrated learning from other behaviours.

pdf document

Reinforcement Learning for Visual Servoing of a Mobile Robot

Proceedings of the Australian Conference on Robotics and Automation (ACRA2000), (Melbourne Australia, August 2000)

This paper applied the algorithm to visual servoing of a mobile robot.

pdf document

short movie (960KB)

Q-Learning in Continuous State and Action Spaces

Proceedings of the 12th Australian Joint Conference on Artificial Intelligence © Springer-Verlag (Sydney Australia, 1999).

This paper demonstrated improved performance through advantage learning. It also contained a survey of continuous state and action q-learning methods and argued for the importance of off-policy learning. If you are interested in this paper, please also consider taking a look at my PhD thesis (above). The thesis is more up to date (better survey) and describes experiments on real robots, rather than just simulations.

pdf document

Autonomous Control and Guidance for an Underwater Robotic Vehicle

Proceedings of the International Conference on Field and Service Robotics (FSR'99) (Pittsburgh, USA, September 1999)

pdf document

Reinforcement Learning for a Visually-Guided Autonomous Underwater Vehicle

Proceedings of the International Symposium on Unmanned Untethered Submersibles Technology (UUST'99) (Durham, New Hampshire, USA, August 1999)

pdf document

Reinforcement Learning applied to the control of an Autonomous Underwater Vehicle

Proceedings of Australian Conference on Robotics and Automation (Brisbane, Australia, March 1999).

This paper introduced the wire fitted neural network algorithm.

pdf document

Development of a Visually-Guided Autonomous Underwater Vehicle

Proceedings of IEEE OCEANS'98 (Nice, France, September 1998)

pdf document