On Problematic Robotic Thresholds

Authors

Adam K. Taylor, University of Nebraska-LincolnFollow

Date of this Version

Fall 12-4-2015

Document Type

Article

Citation

Taylor, Adam K. On Problematic Robotic Thresholds. Thesis. University of Nebraska-Lincoln, 2015. N.p.: n.p., n.d. Print.

Comments

A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Computer Science, Under the Supervision of Sebastian Elbaum. Lincoln, Nebraska: December, 2015

Copyright (c) 2015 Adam K. Taylor

Abstract

Large configuration spaces present difficulties for developers validating large soft- ware systems and for users selecting the proper configuration to achieve the desired runtime behavior. Robot systems face the same challenges as they may have hundreds of configurable parameters. Our work focuses on co-robotic systems, those in which robots and humans work closely together to augment each other’s capabilities. We aim to leverage the user’s knowledge about a system to help determine configuration errors. To accomplish this, users mark runtime failures while observing the system in operation. A marked error indicates the robot “did something when it should not have” or was “not doing something when it should have.” In this thesis we have developed an approach that identifies predicates involving configuration parameters that may be relevant to each error type and can suggest adjustments based on the outcome of those predicates. In this work we present the following 1) A method to statically analyze Python and C++ code to identify threshold predicate comparisons (predicates with values initialized by configura- tion parameters that have an effect on specific execution patterns). 2) A characterization of the configuration space of popular robot systems. 3) A recommendation approach for configuration adjustment that combines user input with program analysis. 4) Three case studies assessing the approach and characterizing threshold predicate comparisons present in a running system.

Adviser: Sebastian Elbaum