"Computational Mechanisms for Action Selection"

How do animals choose appropriate behaviours on a moment-by-moment basis? How can autonomous agents such as robots be hard-wired in such a way as to allow them to always select sensible actions in dynamic and complex domains? How can long-term needs (motivations, goals) be combined with opportunism and threat avoidance (reactive behaviour) to allow for flexible and robust behavioural choice?

Thesis by ftp
Thesis hardcopy by post

refereed papers:
Defining the action selection problem
The use of hierarchies for action selection
An evaluation of Maes' "bottom-up mechanism for action selection"

ANIMALS software (code for simulated environment)

other work on action selection (Phil Goetz's page)

home page

Toby Tyrrell : T.Tyrrell@noc.soton.ac.uk