Post-graduate theses
Current Record: 714 of 824
|
|
| Identifier |
uch.csd.msc//1998aartav |
| Title |
Πλοήγηση ρομποτικού συστήματος σε προκαθορισμένο μονοπάτι |
| Alternative Title |
Navigation of a robotic system on |
| Creator |
Artavanis, Andreas G
|
| Abstract |
Navigation of a robotic system is among the most interesting research topics in the field of robotics, aiming at the development of systems with a high degree of autonomy. Most approaches adopt the use of landmarks, either natural or artificial, or the construction and use of geometric representations of the environment. However, such approaches are based on assumptions that are occasionally not met in practice and their application on real systems may not always be feasible. In this work a methodology is studied for the navigation of an autonomous system on a predefined path, in which we minimize the need for previous knowledge regarding the environment. This methodology best fits to indoor workspaces, making use only of the systems' visual sensors. A two-phase strategy is employed: (a) in the learning phase the system is shown a path, and (b) in the navigation phase the system attempts to track the learned path. The learning phase is a prerequisite to the navigation procedure. During this phase, a teacher interactively directs the system on a specific path, while the system is recording the route as an image sequence which is stored in its memory. During the navigation phase, the system aims at moving on the learned path. In order to accomplish this, the system compares the images acquired at each step of the navigation procedure with the reference images recorded in the learning phase. The differences in position and orientation between the two images are estimated and the result is used towards the correction of the position and the orientation of the system. By repeating the above steps throughout the navigation procedure, the system is not expected to deviate from the predefined path and to move successfully towards its goal. To estimate the position of the system with respect to the reference images, two-dimensional characteristics are extracted from the environment and used. In the current work we studied, implemented and evaluated three methods for the position estimation with respect to the reference images, using two-dimensional characteristics. The first method uses correspondences of vertical lines, the second method uses correspondences of line segments and the third method uses point correspondences between the two images. The results of the position estimation of the system obtained from the above methods are presented and evaluated, using both synthetic and real images.
|
| Issue date |
1998-11-01 |
| Date available |
1998-11-23 |
| Collection
|
School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
|
|
|
Type of Work--Post-graduate theses
|
| Permanent Link |
https://elocus.lib.uoc.gr//dlib/b/7/c/metadata-dlib-1998aartav.tkl
|
| Views |
455 |
Collections
