Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.540224
Title: A biologically inspired neural network for robot navigation
Author: Nichols, Eric James
Awarding Body: University of Ulster
Current Institution: Ulster University
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Abstract:
This research implements two spiking neural network (SNN) based robot navigation systems inspired by the sensory fusion and control structures of biological systems. The work builds on an in-depth review of robotic and SNN systems. The robotic review points to perception, cognition and reasoning as intelligent attributes that are lacking in modern robotic systems but are performed effortlessly by the nervous system within biologic organisms. The SNN review focusses on models of synapses and neurons and the associated architectures that use long and short term plasticity rules as mechanisms for learning. This review also focusses on self-organisation and there is a brief description of the peripheral and central nervous systems. Two self-organising SNN architectures are presented and their performances are verified on wall following tasks experimentally. Input is obtained from infrared sensors in the first SNN. Structured learning maps the input to appropriate output by self-organising the SNN architecture as the robot experiences novel environments. Information is routed through the SNN with biological inspiration from synapses with short term plasticity. Working memory is implemented by fusing prior and current conditions to provide a richer sense of the environment. Learning occurs online in a supervised manner using hand-crafted rules. The second SNN fuses inputs from laser and sonar sensors. The self-organising structure from the first SNN is maintained and the biological precision of short term plasticity is increased using a facilitating synaptic model at every synapse until the final layer where depressing synapses are used. Long-term synaptic plasticity is implemented online using the temporal difference learning rule to enable the robot to learn to associate the correct movement with the appropriate input conditions. Results of experiments on each SNN are presented and conclusions are drawn. vii
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.540224  DOI: Not available
Share: