On Robotic Work-Space Sensing and Control
(2013) In PhD Thesis TFRT-1098- Abstract
- Industrial robots are fast and accurate when working with known objects at precise locations in well-structured manufacturing environments, as done in the classical automation setting. In one sense, limited use of sensors leaves robots blind and numb, unaware of what is happening in their surroundings. Whereas equipping a system with sensors has the potential to add new functionality and increase the set of uncertainties a robot can handle, it is not as simple as that. Often it is difficult to interpret the measurements and use them to draw necessary conclusions about the state of the work space. For effective sensor-based control, it is necessary to both understand the sensor data and to know how to act on it, giving the robot... (More)
- Industrial robots are fast and accurate when working with known objects at precise locations in well-structured manufacturing environments, as done in the classical automation setting. In one sense, limited use of sensors leaves robots blind and numb, unaware of what is happening in their surroundings. Whereas equipping a system with sensors has the potential to add new functionality and increase the set of uncertainties a robot can handle, it is not as simple as that. Often it is difficult to interpret the measurements and use them to draw necessary conclusions about the state of the work space. For effective sensor-based control, it is necessary to both understand the sensor data and to know how to act on it, giving the robot perception-action capabilities.
This thesis presents research on how sensors and estimation techniques can be used in robot control. The suggested methods are theoretically analyzed and evaluated with a large focus on experimental verification in real-time settings.
One application class treated is the ability to react fast and accurately to events detected by vision, which is demonstrated by the realization of a ball-catching robot. A new approach is proposed for performing high-speed color-based image analysis that is robust to varying illumination conditions and motion blur. Furthermore, a method for object tracking is presented along with a novel way of Kalman-filter initialization that can handle initial-state estimates with infinite variance.
A second application class treated is robotic assembly using force control. A study of two assembly scenarios is presented, investigating the possibility of using force-controlled assembly in industrial robotics. Two new approaches for robotic contact-force estimation without any force sensor are presented and validated in assembly operations.
The treated topics represent some of the challenges in sensor-based robot control, and it is demonstrated how they can be used to extend the functionality of industrial robots. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/4124506
- author
- Linderoth, Magnus LU
- supervisor
- opponent
-
- Sznaier, Mario, Electrical and Computer Engineering Department, Northeastern University, Boston, USA
- organization
- publishing date
- 2013
- type
- Thesis
- publication status
- published
- subject
- keywords
- Robotics, computer vision, color constancy, motion blur, filter initialization, visual tracking, on-line trajectory generation, ball-catching robot, force control, robotic assembly, force estimation, perception-action capabilities.
- in
- PhD Thesis TFRT-1098
- pages
- 219 pages
- publisher
- Department of Automatic Control, Lund Institute of Technology, Lund University
- defense location
- Lecture hall M:B, M-building, Ole Römers väg 1, Lund University Faculty of Engineering
- defense date
- 2013-11-22 11:15:00
- ISSN
- 0280-5316
- 0280-5316
- ISBN
- 978-91-7473-670-0
- 978-91-7473-669-4
- project
- ROSETTA
- language
- English
- LU publication?
- yes
- id
- ccac0d12-9040-4acb-82fb-699408a95742 (old id 4124506)
- date added to LUP
- 2016-04-01 14:09:31
- date last changed
- 2019-05-23 16:07:37
@phdthesis{ccac0d12-9040-4acb-82fb-699408a95742,
abstract = {{Industrial robots are fast and accurate when working with known objects at precise locations in well-structured manufacturing environments, as done in the classical automation setting. In one sense, limited use of sensors leaves robots blind and numb, unaware of what is happening in their surroundings. Whereas equipping a system with sensors has the potential to add new functionality and increase the set of uncertainties a robot can handle, it is not as simple as that. Often it is difficult to interpret the measurements and use them to draw necessary conclusions about the state of the work space. For effective sensor-based control, it is necessary to both understand the sensor data and to know how to act on it, giving the robot perception-action capabilities.<br/><br>
<br/><br>
This thesis presents research on how sensors and estimation techniques can be used in robot control. The suggested methods are theoretically analyzed and evaluated with a large focus on experimental verification in real-time settings.<br/><br>
<br/><br>
One application class treated is the ability to react fast and accurately to events detected by vision, which is demonstrated by the realization of a ball-catching robot. A new approach is proposed for performing high-speed color-based image analysis that is robust to varying illumination conditions and motion blur. Furthermore, a method for object tracking is presented along with a novel way of Kalman-filter initialization that can handle initial-state estimates with infinite variance. <br/><br>
<br/><br>
A second application class treated is robotic assembly using force control. A study of two assembly scenarios is presented, investigating the possibility of using force-controlled assembly in industrial robotics. Two new approaches for robotic contact-force estimation without any force sensor are presented and validated in assembly operations.<br/><br>
<br/><br>
The treated topics represent some of the challenges in sensor-based robot control, and it is demonstrated how they can be used to extend the functionality of industrial robots.}},
author = {{Linderoth, Magnus}},
isbn = {{978-91-7473-670-0}},
issn = {{0280-5316}},
keywords = {{Robotics; computer vision; color constancy; motion blur; filter initialization; visual tracking; on-line trajectory generation; ball-catching robot; force control; robotic assembly; force estimation; perception-action capabilities.}},
language = {{eng}},
publisher = {{Department of Automatic Control, Lund Institute of Technology, Lund University}},
school = {{Lund University}},
series = {{PhD Thesis TFRT-1098}},
title = {{On Robotic Work-Space Sensing and Control}},
url = {{https://lup.lub.lu.se/search/files/3819777/4124507.pdf}},
year = {{2013}},
}