Use this URL to cite or link to this record in EThOS:
Title: Computational Neuroethology using Programmable Logic devices
Author: Pearson , Martin James
ISNI:       0000 0004 2668 878X
Awarding Body: University of the West of England
Current Institution: University of the West of England, Bristol
Date of Award: 2007
Availability of Full Text:
Access from EThOS:
This thesis focuses on the use of programmable digital logic hardware to implement biologically plausible models of neural systems to substantiate neuroethological hypotheses in real-time using 'mobile robotics. The author contends that, in order to model the immense complexity of sections of the central nervous in real-time, using bounded computational resources, one must address the level of model abstraction that is adopted. Further, the natural dynamic behaviour of each component structure of the overall system model must be assessed, and an appropriate processing strategy based on those observations should be adopted. This contention has been substantiated by the implementation of a biologically inspired model of the rodent whisker sensory system and its embodiment onto a mobile robotic platform that can perform ethologically plausible behaviour. Programmable digital logic has been adopted because of its commercial availability, re-usability and future scalability; all three contributing to the argument that reconfigurable digital hardware is an under utilised technology in the field of computational neuroscience, especially in the area of embodied computational neuroscience, or computational neuroethology. The first novel contribution to this field is a suite of neural processing cores developed for Field Programmable Gate Arrays (FPGA), that can reproduce large, hardware implemented, networks of biologically plausible spiking· neurons proposed by neuroscientists using conventional software based neuronal network simulators. These processing cores differ from similar hardware neural processors in that they can maintain a fixed update period independent of the activity of the neural network model itself, and that multiple cores can be cascaded together to support larger networks with no loss of temporal performance. The second novel contribution is the design of a control architecture that can integrate these hardware based processing cores with software based models in aPC to form heterogeneous processing networks. This allowed models of disparate regions of the rodent brain to be implemented using processing strategies best suited to reproduce the behavioural characteristics of each region and maintain a real-time system wide performance. This control architecture was demonstrated on a mobile robotic platform in experiments to test current neuroethological hypotheses ofthe rat whisker sensory system.
Supervisor: Not available Sponsor: Not available
Qualification Name: Not available Qualification Level: Doctoral
EThOS ID:  DOI: Not available