Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Learning Task-Based Robotic Grasping with Vision, Haptics and Proprioception.
KTH, School of Computer Science and Communication (CSC).
2012 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

Grasping is an essential capability of service robots in order to interact with the environment. To perform a grasp that satisfies a designated task is challenging since there are many constraints associated with the task. This work studies the topic of the robot grasp adaptation with regard to both geometrical and stability requirements through a supervised learning approach. By giving the geometrical task and stability labels to the grasps that the robot has implemented during the self-exploration, the robot learns to classify the task affordance of a novel grasp given the sensory observations.

Gaussian Mixture Model probability modeling and Support Vector Machine are evaluated as the classifiers given one-shot tactile information. And for the temporal sensory information, the classifiers are chosen as Gaussian Mixture Regression and Support Vector Machine on polynomial regression coefficients. Bayesian network, as a generic model is also evaluated for its performance on both one-shot and temporal sensory information. The result shows that one-shot data lead to higher stability prediction accuracy for the discriminative models such as Support Vector Machine, while Bayesian network is more flexible in predicting grasp success even without tactile sensory observations.

Abstract [sv]

Att greppa är en väsentlig förmåga av servicerobotar för att interagerar med miljön. För att utföra en grepp som uppfyller en bestämd uppgift är utmaning för det finns många begränsningar i samband med uppgiften. Detta arbete studerar ämnet att robot grepp anpassningen vad gäller båda geometriska och stabilitet krav genom övervakad lärande. Genom att tilldela geometriska uppgiften och stabilitet etiketter till griperna att roboten har genomförts under den själv-prospektering, roboten lär att klassificera uppgiften av en ny grepp med tanke på de sensoriska observationer.

Gaussisk Blandning Modell sannolikhet modellering och Stödvektormaskin bedöms som klassificerare ges enda-bildruta taktil information. Och för den tidsmässiga sensorisk information, klassificerare är valdes som Gaussisk Blandning Regression och Stödvektormaskin för polynom regressionskoefficienter. Bayesianska nätverk, som en generisk modell också utvärderas för dess prestanda på båda enda-bildruta och tidsmässiga sensorisk information. Resultatet visar att enda-bildruta data ledar till högre stabilitet prognos noggrannhet för de diskriminerande modeller som Stödvektormaskin, men Bayesianska nätverk är mer flexibel i att förutsäga grepp framgång även utan taktila sensoriska observationer.

Place, publisher, year, edition, pages
2012.
Series
Trita-CSC-E, ISSN 1653-5715 ; 2012:039
National Category
Computer Science
Identifiers
URN: urn:nbn:se:kth:diva-130995OAI: oai:DiVA.org:kth-130995DiVA: diva2:654441
Educational program
Master of Science - Software Engineering of Distributed Systems
Uppsok
Technology
Supervisors
Examiners
Available from: 2013-10-07 Created: 2013-10-07

Open Access in DiVA

No full text

Other links

http://www.nada.kth.se/utbildning/grukth/exjobb/rapportlistor/2012/rapporter12/wang_lu_12039.pdf
By organisation
School of Computer Science and Communication (CSC)
Computer Science

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 30 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf