Dissertation Title: Precise Assembly of Truss Structures by Distributed Robots

Date:  November 7th 2014

Time:  9:30 am

Venue:  ECOT 831

 

Committee:

Nikolaus Correll, (Chair)

Sriram Sankaranarayanan

Tom Yeh

Eric Frew

Daniel Scheeres

 

Abstract:

 

Assembly robots have been in operation in industry for decades.  However, in the field, robotic assembly has no short-term feasible applications.  The space industry desires large space telescopes, but they require micron-level structural precision.  Previous attempts were ruled out because the precisely machined components were expensive, heavy, and prone to failure.  In this thesis, I describe the Intelligent Precision Jigging paradigm, a solution to the problem of practical robotic assembly.  Intelligent Precision Jigging Robots (IPJRs) precisely position commodity parts, enabling coarse external manipulators to weld and perform other tasks. I present algorithms for optimizing assembly sequences and for implementing Simultaneous Localization and Mapping (SLAM) to maintain an estimate of the assembly process using local strut length measurements.  I define a truss probability model and a minimizing metric based on the covariance trace.  I show that structure error grows cubically with node count when error correction is not used.  I then present an approach for generating near-optimal assembly sequences: start assembly on a central location of the structure, greedily assemble to minimize error, and swap assembly steps until a local minimum is found.  I show that this method consistently generates better sequences than any process alone. I then describe and simulate SLAM with four estimators; Linear Least Squares, Extended Kalman Filter, Unscented Kalman Filter, and Maximum Likelihood Estimator.  I show that the Maximum Likelihood Estimator is the best and that SLAM mitigates the growth of structure error. Finally, I present the results of physical assembly trials on a telescope truss made of aluminum tubes, assembled by three IPJRs using two methods: an open loop approach, and an MLE-SLAM approach.  I show that the MLE-SLAM assembly algorithm works even when the physical trials included unmodeled processes such as deformation under gravity.

 

BIOGRAPHICAL NOTES

 

Erik Komendera was awarded a BSE in Aerospace Engineering in 2007 by the University of Michigan, and was awarded a MS in Computer Science in 2012 by the University of Colorado.  His research interests include robotic assembly, state estimation, control systems, and chaotic dynamics.  He will join NASA at Langley Research Center as a roboticist.

 

Leave a Reply

Set your Twitter account name in your settings to use the TwitterBar Section.