Data collection for screwdriving - Robotics Institute Carnegie Mellon University

Data collection for screwdriving

Reuben M. Aronson, Ankit Bhatia, Zhenzhong Jia, and Matthew T. Mason
Workshop Paper, RSS '17 Workshop on (Empirically) Data-Driven Manipulation, July, 2017

Abstract

We built an instrumented screwdriver stage to detect categories of failures not previously documented and to record their sensor information for future prediction. Our data collection process was arduous, relying on manual resets and data labeling, but the low incidence of failure in the operation required a large number of runs. Furthermore, the data collection process itself damaged the screws and affected the results. We intend to design another collection process with more automation to gather more and better data.

BibTeX

@workshop{Aronson-2017-126625,
author = {Reuben M. Aronson and Ankit Bhatia and Zhenzhong Jia and Matthew T. Mason},
title = {Data collection for screwdriving},
booktitle = {Proceedings of RSS '17 Workshop on (Empirically) Data-Driven Manipulation},
year = {2017},
month = {July},
}