11 October 2010

Prof Kevin Gurney wins EPSRC grant

Kevin Gurney is part of a successful EPSRC award for a project entitled "Dual Process Control Models in the Brain and Machines with Application to Autonomous Vehicle Control" The project is a spin-off from REVERB, a previous EPSRC award led by Kevin. The total value of the award is ~£340K, and ~£60K comes to Sheffield (mainly Kevin's time). A brief description of the research is below.

The fundamental computational problems faced by autonomous agents and intelligent controllers are present irrespective of their mode of implementation - whether it be biological or artificial. Therefore, given that biology appears to have solved many of these problems, one expedient strategy for developing artificial agents is to reverse engineer their biological solution. Here, we propose to deploy this approach by harnessing similarities we have observed between specific features of the architecture of the vertebrate brain, and a modular class of engineering control systems. The proposal builds on prior work by Prof Gurney in modelling the relevant brain architectures, and Dr Hussain in developing the intelligent control systems in question and applying them to autonomous vehicle control (AVC). In particular, this project aims to leverage new insights gained on the EPSRC-funded REVERB project (#EP/C516303/1) led by Prof Gurney. The research will have immediate impact in AVC, promises wider ramifications in control theory, and will yield novel hypotheses concerning functional principles at work in the brain.

...More specifically,

(1) We will focus on the concepts of automatised and controlled (or executive) processing in the brain and how they might map onto modular control solutions.

(2) Our forcing domain is autonomous vehicle control (AVC) with an existing proven design, whose architecture has been shown to have similarities with neural and psychological models of executive and automatised control.

(3) We therefore aim to build and validate a new generation real-time AVC controller, more directly inspired by the biological ideas, and to evaluate the benefits of such a controller using realistic validated vehicle models (from the project's industrial collaborators) over existing conventional adaptive controller methods.

(4) In turn, by implementing the general principles of dual process theory in our forcing domain, we aim to discover new general mechanisms supporting controlled and automatic processing that will lead to novel hypotheses as to how these modes of operation are supported in the brain.