Evolutionary design of digital trajectory-tracking controllers for robotic manipulators

Allaoui, C

Abstract

The design of digital trajectory-tracking controllers for robotic manipulators is a challenging task, since such manipulators are multivariable non-linear plants. In addition, in many applications of robotic manipulators, it is required that very high-accuracy trajectory-tracking performance be achievable even in the presence of unpredictable payload variations. These requirements can all be met to some extent by application of the previously developed fast-sampling digital PID controllers to robotic manipulators. Indeed, for such controllers, it is possible to prove a series of very reassuring robustness results using only the Markov parameters associated with locally linearised representations of robotic manipulators.
However, these theoretical optimisation results for digital PID controllers are only valid as sampling periods become vanishingly small. In practice, of course, the sampling periods of digital controllers remain non-zero; but, in such cases, no theoretical optimisation results are available. There is, therefore, a great need for some alternative optimisation procedure that will facilitate the non- asymptotic design of digital PID controllers for robotic manipulators.
This design need is addressed in this thesis. In particular, the following evolutionary optimisation techniques are used to design digital trajectorytracking controllers for robotic manipulators:
(i) genetic algorithms,
(ii) non-adaptive evolution strategies
(iii) adaptive evolution strategies.

It is shown that, with increasing effectiveness, these techniques are very useful in the design of high-accuracy digital PID controllers. These techniques are illustrated by the presentation of simulation results for a typical three-link robotic manipulator performing a range of demanding trajectory-tracking tasks in the presence of unpredictable payload variations. In addition, these evolutionary optimisation techniques are also used in the design of unconstrained digital PID controllers, in which all elements of the controller matrices are used as the design parameters.
In order to validate these evolutionary design techniques in practice, an experimental laboratory investigation is also undertaken. This involves the practical implementation, in the case of a direct-drive two-link robotic manipulator, of digital PID trajectory-tracking controllers designed using evolutionary techniques. The results thus obtained indicate that such optimisation techniques greatly facilitate the tuning of digital PID controllers for robotic manipulators under practical non-asymptotic conditions.

Citation

Allaoui, C. Evolutionary design of digital trajectory-tracking controllers for robotic manipulators. (Thesis). University of Salford

• Thesis Type: Thesis
• Deposit Date: Dec 22, 2021
• Publicly Available Date: Dec 22, 2021
• Publisher URL: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.299204
• Related Public URLs: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.299204
• Additional Information: Funders : The British Council/Foreign and Commonwealth Office;The Algerian Foreign Office
• Award Date: Nov 1, 1998