US mini logoHome | A-Z Index | People | Reference | Contact us
University of Sussex
Home | About | Research | People | Publications | Talks| Links | Local

CCNR People

Robert Vickerstaff - Alumnus

Phone
Location
Email
Webpage

Supervisors:
++44 (1273) 87-2948
JMS 3D8
robertvi@sussex.ac.uk
http://www.ladysmith.eclipse.co.uk/

Ezequiel Di Paolo, Micheal O'Shea
Search the CCNR bibliography for papers by Robert Vickerstaff

Research Interests

I am currently using artificial evolution to produce simulated agents capable of the navigation technique known as path integration (P.I.). P.I. (or "dead reckoning" navigation) is used by many species, including insects and mammals, as a foundation of their navigational strategies. It can operate in the absence of landmarks and requires only knowledge of compass bearing and the distance travelled.

The task of evolving P.I. attracted me because it is both non-trivial and potentially informative to our understanding of neuroscience and animal behaviour. I wanted a task hard enough to test the properties of different classes a neural model, by way of direct comparison on the same task.

The neural mechanisms underlying P.I. are currently poorly understood and to date only hand designed neural implementations have been proposed. The analysis of neurally controlled evolved agents might therefore provide informative alternative models, since (a) they would not be constrained by our ability to design neural mechanisms by hand and (b) they would have been produced by an evolutionary process.

See also

Site maintained by: Marieke Rohde Disclaimer | Feedback