Nekalibrirano vodenje robota s pomočjo računalniškega vida

doktorsko delo

Rok Klobučar (Author), Riko Šafarič (Mentor), Miran Brezočnik (Co-mentor)

Abstract

Raziskave s področja vodenja robota s pomočjo računalniškega vida so zelo pomembne. V doktorski disertaciji je opisano vodenje robotskega manipulatorja.Za vodenje s pomočjo umetnega vida je uporabljena polno povezana usmerjena nevronska mreža. Uporabljena je povsem nova arhitektura nevronske mreže in nov algoritem nevronske mreže. Nobeno poznavanje kinematike robota in kalibracUo kamere ni potrebna. Nevronska mreža se uči s pomočjo vrha robota. Po učenju smo testirali natančnost nevronske mreže. Nevronska mreža je morala generirati notranje koordinate glede na zunanjo trajektorijo. Eksperiment je bil izveden na paralelnem manipulatorju z dvema prostostnima stopnjama. Pokazali smo, da lahko s pomočjo taksne arhitekture nevronske mreže robota opravlja zadane naloge. V nalogi je predstavljena aplikacija nevronske mreže za ocenitev nelinearne transformacye med vrhom robota v sliki in notranjih koordinat. Eksperimentalni primeri potrjujejo zanesljivost predlagane metode. Eksperimentalne rezultate smo primerjali s podobno metodo imenovano Broydenova metoda za nekalibrirano vodenje robota s pomočjo umetnega vida.

Keywords

roboti;računalniški vid;vizualno vodenje;usmerjene nevronske mreže;

Data

Language:	Slovenian
Year of publishing:	2008
Source:	[Maribor
Typology:	2.08 - Doctoral Dissertation
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	R. Klobučar]
UDC:	681.532.1(043.3)
COBISS:	243113984
Views:	3421
Downloads:	334
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Uncalibrated visual servo control
Secondary abstract:	Research into robotics visual servo systems is an important content in the robotics field. This dissertation describes a control approach for a robotics manipulator. In this dissertation, a multilayer feedforward network is applied to a robot visual servo control problem. The model uses new neural network architecture and a new algorithm for modifying neural connection strength. No a-prior knowledge is required of robot kinematics and camera calibration. The network is trained using an end-effector position. After training, performance is measured by having the network generate joint-angles for arbitrary end effector trajectories. A 2-degrees-of-freedom (DOF) parallel manipulator was used for the study. It was discovered that neural networks provide a simple and effective way of controlling robotic tasks. This disertation explores the application of a neural network for approximating nonlinear transformation relating to the robot's tip-position, from the image coordinates to its joint coordinates. Real experimental examples are given to illustrate the significance of this method. Experimental results are compared with a similar method called the Broyden method, for uncalibrated visual servo-control.
Secondary keywords:	robots;computer vision;visual servoing;feedforward neural network;
URN:	URN:SI:UM:
Type (COBISS):	Dissertation
Thesis comment:	Univ. v Mariboru, Fak. za strojništvo
Pages:	XVIII, 112 f.
Keywords (UDC):	applied sciences;medicine;technology;uporabne znanosti;medicina;tehnika;industries;crafts and trades for finished or assembled articles;industrije;obrti in rokodelstva za sestavljanje in dodelavo izdelkov;precision mechanisms and instruments;fina mehanika in instrumenti;automatic control technology;smart technology;avtomatska kontrola;inteligentni sistemi;
ID:	985488