Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.538953
Title: Biologically inspired sensory processing for mobile robots using Spiking Neural Networks
Author: McBride, Michael F.
Awarding Body: University of Ulster
Current Institution: Ulster University
Date of Award: 2010
Availability of Full Text:
Access from EThOS:
Abstract:
This thesis is focused on research into biologically inspired sensory fusion for a mobile robot. The approach is based upon a bio-inspired Liquid State Machine (Reservoir Computing) paradigm, utilising Spiking Neural Networks in the reservoir as the core of the sensory fusion system, with a conventional classical artificial neural network in the readout phase. The connectivity and structure of the LSM is inspired by the biological example of the mammalian brain and in particular by the connectivity of the somatosensory cortex. The use of the reservoir computing paradigm allows for effective integration of data from different sensory modalities within the reservoir and permits snapshots of the internal state to be captured for subsequent processing. The use of such an approach provides a novel method for autonomous systems to combine information, in a method which is more closely inspired by nature. The experimental analysis of this research investigates a robot traversing an environment using multiple sensory inputs from multiple sensor types and experiencing varying sensory conditions. The research investigates parameters for sensor data coding and creating a LSM for processing sensor information. An LSM structure is presented to combine the sensor information within its structure. The empirical assessment of the LSM sensor fusion experiments of the robot obstacle avoidance is presented. The experiments demonstrate how the fusing of separate sensor data in the LSM improves the performance of the robot over the performance of processing a single sensor type in the LSM
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.538953  DOI: Not available
Share: