Evolutionary algorithms for practical sensor fault tolerant control

• Main Author: Hirayama, Yoshikazu
• Published: University of York 2007
• Subjects: 003.5
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.485096

The Shaky Hand is a multi-input, multi-output laboratory demonstrator which is modelled on a village fete game. In the original, the aim is to guide, by hand, a wire loop along a wire which has been bent to form a meandering track, 'without touching the loop to the wire. In the original game, touching the hand-held loop against the wire track sets off a loud warning bell and the player loses. The thesis presents the research work associated with the quest for practical solutions to a generic problem: the correct operation of a fallible system. The work covers three distinct areas: modelling of the demonstrator, design and construction of a physical system, and evoiution of algorithms for control of the demonstrator in practice in the presence of sensor faults, using Cartesian Genetic Programming (CGP). The third area forms the core of the thesis. The key challenges in creating the virtual environment to train for generic sensor fault tolerant algorithms are considered and addressed. The evolved algorithms are analysed and then verified using the demonstrator in practice. The practical results showed that sensor fault tolerant control was successfully achieved.