Estimation of objects’ inertial parameters, and their usage in robot grasping and manipulation

Mavrakis, Nikolaos ORCID: 0000-0002-3138-2633 (2020). Estimation of objects’ inertial parameters, and their usage in robot grasping and manipulation. University of Birmingham. Ph.D.

Preview

Mavrakis2020PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (16MB) | Preview

Abstract

The subject of this thesis is the estimation of an object's inertial parameters by a robotic arm, and the exploitation of those parameters in the design of efficient manipulation criteria. The inertial parameters of objects describe the resistance of the object to an applied force, and dictate its motion. Research has shown that humans intuitively exploit them for their everyday manipulations. As humans are very capable of performing efficient manipulations, it is natural that robots should use the inertial parameters as well. Additionally, as the inertial parameters are not straightforward to calculate, there is the need for development of methods that can estimate them online.
This thesis focuses on two directions, developing novel methods so that robots can accurately estimate the inertial parameters of an object, as well as developing manipulation criteria that can make robot task completion more efficient. The relevant literature is gathered, categorised and analytically described, and the innovation gaps are identified. The thesis offers novel research solutions on the problem of estimation of the inertial parameters with minimal robot interaction. The paradigm is shifted from the existing literature, and a data-driven estimation algorithm is introduced, that achieves accurate results with both simulated and real data. Additionally, the presented research is offering novel manipulation criteria that are affected by the object's inertial parameters. The results suggest that knowledge of the inertial parameters can make the robot tasks more power-efficient and safe to their surroundings. The core methodology is shown to be versatile to the robotic platform. Though most experiments are performed on a terrestrial robot, a numerical example is also shown for a space robot.
The results of the thesis suggest that the developed methods can be used in various environments, with the most suitable being extreme environments where accuracy, efficiency and autonomy is required.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):