Abstracts

Modeling, estimation and control for serial flexible robot arms

by Arto Kivila




Institution: Georgia Tech
Department:
Year: 2017
Keywords: Multi-body dynamics; Extended Kalman filter; Input-shaping; Non-linear control
Posted: 02/01/2018
Record ID: 2222111
Full text PDF: http://hdl.handle.net/1853/58644


Abstract

Industry demands for high precision automation equipment have led to heavy, stiff and therefore expensive, inefficient, and potentially dangerous serial link manipulators. Industry has been reluctant to move towards lighter, and therefore inherently more flexible manipulators, despite the potential benefits of lower costs, increased throughput and improved safety. Advancements in data processing capabilities, sensing capabilities and control theory during the past couple of decades can potentially solve the perceived problems with flexible serial manipulators. In general, there is no such thing as a fully rigid manipulator: even current industrial robots exhibit small flexibilities. In addition, there are applications such as space-robotics and nuclear waste tank inspection/cleaning jobs where light and long links are the only option. Most research in the past has focused on single link manipulators and planar robot arms rather than spatial multi-link robots. This dissertation presents a systematic approach for obtaining natural frequencies and mode-shapes for n-link spatial serial structures based on transfer matrices. The method is validated with experiments and software simulations. A low-order dynamical model for n-link flexible manipulators in spatial configurations is presented. The model is verified with finite element simulations, and hardware experiments. The low-order model is the basis of an extended Kalman filter based estimator that allows sensor-based predictions of the flexible states. Accelerometer and strain gage based feedback is examined. Accelerometer based feedback is verified with experiments. In order to damp out he oscillations multi-link flexible arms caused by the reference command, an optimized input shaping algorithm for multiple frequency ranges is presented. The results are confirmed with FEA analysis and experiments. The controllability of natural modes is discussed and analyzed. An inversion based closed-loop controller is presented that guarantees stable joint trajectory tracking for flexible manipulator arms. A singular pertubation based controller is presented to actively damp out the vibrations in the arm. A test bed that provided verification of the claims made in this dissertation was designed and constructed. The test bed has 3 actuators and 2 flexible links.Advisors/Committee Members: Singhose, William E. (advisor), Book, Wayne J. (committee member), Sadegh, Nader (committee member), Ueda, Jun (committee member), Vaughan, Joshua (committee member).