Avoidance filtering for human-commanded manipulation systems using constrained control

Michael Kearney (2010). Avoidance filtering for human-commanded manipulation systems using constrained control PhD Thesis, School of Mechanical and Mining Engineering, The University of Queensland.

       
Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
s40274982_phd_abstract.pdf Final Thesis Abstract application/pdf 60.55KB 3
s40274982_phd_totalthesis.pdf Final Thesis Lodgement application/pdf 3.54MB 19
Author Michael Kearney
Thesis Title Avoidance filtering for human-commanded manipulation systems using constrained control
School, Centre or Institute School of Mechanical and Mining Engineering
Institution The University of Queensland
Publication date 2010-05
Thesis type PhD Thesis
Total pages 243
Total colour pages 40
Total black and white pages 203
Subjects 09 Engineering
Abstract/Summary In the operation of many human-commanded machines it would be advantageous to have a control technology that assists the operator avoid collisions with obstacles. Electric mining shovels are a good candidate. These machine have slow dynamics requiring significant look-ahead by the operator to steer around obstacles. Should collision occur, with say a haul truck, the momentum of the machine is transformed into large damage-causing impact-forces. And because the field of view of the operator is often occluded, the operator may not be able to see obstacles (e.g. trucks) that need to be avoided. This thesis is concerned with the problem of determining how to alter the operator's command of such a machine to automatically avoid collision. A solution to this problem is termed an avoidance filter. Two avoidance filtering strategies are developed: (i) the receding horizon avoidance filter (RHAF) based on ideas of constrained model predictive control and (ii) the invariant set avoidance filter (ISAF) based on ideas from set-theoretic control. These two approaches are at different ends of the spectrum of constrained control methods and an aim of the work is to understand which is best suited to the problem. The RHAF adapts the model predictive control framework to solve a finite-horizon avoidance filtering problem at each time step. This problem is posed as a mixed-integer program with the constrained linear system dynamics and obstacles represented as inequalities. The thesis develops and evaluates a formulation of this approach focussing on how to efficiently represent obstacles within the mixed integer program. The ISAF uses set invariance properties to ensure that the state evolution of the system remains within a control invariant subset of the obstacle-free region of the state space. The ISAF algorithm is split into two parts: (i) offline calculation, where the appropriate control invariant subset is determined and (ii) online control where the operator command is altered to ensure that it remains within the control invariant set. This has the advantage of moving much of the computational effort offline. When the human-controlled system is described by a constrained linear model and obstacles are convex polytopes, the filtered command can be determined by the solving a linear or quadratic program for each polytope comprising the invariant set. Extensions to the ISAF are also developed to deal with disturbances and systems where one degree of freedom is periodic in position. The RHAF and the ISAF are evaluated experimentally through implementation on an electric mining shovel. Both algorithms were found to work. Present computing technologies limit the size of the problem that can be solve and this is a practical barrier to the pursuit of these ideas. However, the work provides a foundation on which future implementations, less restricted by computer speed, can be developed.
Keyword Model Predictive Control
Obstacle Avoidance
Set-Theoretic Control
Constrained Control
Operator Assistance
Additional Notes Colour pages: 29 30 31 67 70 72 73 75 76 79 83 85 86 95 98 99 100 111 113 114 131 142 143 144 146-148 150 176 178 179 189 192-199

 
Citation counts: Google Scholar Search Google Scholar
Access Statistics: 148 Abstract Views, 22 File Downloads  -  Detailed Statistics
Created: Mon, 22 Nov 2010, 16:01:17 EST by Mr Michael Kearney on behalf of Library - Information Access Service