© University of Kent - Contact | Feedback | Legal | Cookies
The University of Kent, Canterbury, Kent, CT2 7NZ, T +44 (0)1227 764000
Mr Samuel Perkins
Research Student
|
|
|
Publications
My publications are available from the University of Kent's Academic Repository.
The title of my M.Sc. thesis was "Hypervelocity Impact Detection: An Investigation into Piezoelectric Response of PVDF Films". It contains experimentally collected data regarding the practical application of poly(vinylidene fluoride) as the functional interface of impact sensors that were used to detect hypervelocity impacts of projectiles fired from a light gas gun.
PhD Project Summary
Research for my Ph.D. is focussed upon amorphous computing. I have a particular interest in the effects that different types of noise may have upon the effectiveness of algorithms designed to create the emergence of self-organising behaviour.
Research Interests
I am a member of the following research groups:
My main interest is in combining amorphous computing and robotics to allow a distributed artificial intelligence to gain autonomous control of a single robot. The emergence of self-organising behaviour across an amorphous computer can be used to control a robot in real time, and to store the spatially computed results of attempted robot actions. In this way, it may be possible to allow a robot to repeat actions it has already successfully accomplished without having to wait for a spatial result to be re-computed. In other words, undertake actions more efficiently with experience, with the possibility that experience can be cumulative. This interests me for two main reasons:
1. With the use of an amorphous computer, a robot can be made to perform relatively complex tasks while having autonomy that isn't based on a stored program. The robot is controlled by emergent behaviour across the computer, and the emergence is brought about by a series of random interactions between computing elements: - An emergent autonomy.
2. Since it may be possible to store spatially computed results of attempted robot actions, the distributed AI of the robot may be able to learn how to repeat tasks more efficiently. Since each spatially computed result across the computer is brought about by random interactions, and the quantity of elements in the computer means that each specific result would be brought about by a practically unique set of interactions, then unique patterns of memorised data pertaining to all of the robot's attempted actions and collective experience could emerge. However, no data pertaining to such would ever be stored explicitly: - It is simply an emerged pattern of data, distributed spatially. The translation of the emergent patterns of data would itself be an emergent behaviour.
So, not only could the method of control be an emergent phenomenon, but also the method of learning and storing memory. This is my interest in robotics.
As an undergraduate, I read Physics with Space Science and Systems at the University of Kent. Following this, I read for an M.Sc. by Research and Thesis within the same department. I am now a member of the Computational Intelligence Group within the School of Computing, and I am reading for a Ph.D.
During my time at the university, I have been the team captain of the University of Kent Archery Club, and I was a sports scholar between the years of 2005 and 2010. I have represented the university at a national and international level.