@InProceedings{Petkos2007,
  Title                    = {Load estimation and control using learned dynamics models},
  Author                   = {Petkos, G. and Vijayakumar, S.},
  Booktitle                = {Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems},
  Year                     = {2007},
  Pages                    = {1527--1532},

  Abstract                 = {Classic adaptive control methods for handling varying <span class='snippet'>loads</span> rely on an analytically derived model of the robot's dynamics. However, in many situations, it is not feasible or easy to obtain an accurate analytic model of the robot's dynamics. An alternative to analytically deriving the dynamics is learning the dynamics from movement data. This paper describes a <span class='snippet'>load</span> <span class='snippet'>estimation</span> technique that uses the learned instead of analytically derived dynamics. We study examples where the various inertial parameters of the <span class='snippet'>load</span> are estimated from the learned models, their effectiveness in control is evaluated along with their robustness in light of imperfect, intermediate dynamic models.},
  Doi                      = {10.1109/IROS.2007.4399373},
  Review                   = {- How can we use adaptive control to pick up loads? Picking up things changes dynamics of the system. Two ways of this happening:
- q, dq, ddq sensing: special case of parameter estimation. The whole 10 parameters thing. 
 - however, not all parameters can be identified
 - some params are not excited, while others are linearly dependent
 - could get equations analytically, but often complicated
 - so we'll numerically learn it instead
- another way to do this is to look at torque sensing and map input torque to model
 - need to use a PD controller, then switched a "composite" controller (PD + model)

- so the paper combines these to figure out load weight
 - so convert a nx10*n param system to a 11 model system (10 for the last link, 1 for the rest of the robot)
- estimates model accuracy by looking at the ratio between fdfwd (which the model generates) term and fdbk load},
  Timestamp                = {2011.02.03}
}

{"_id":"d4axdSmtP7fTYXumj","bibbaseid":"petkos-vijayakumar-loadestimationandcontrolusinglearneddynamicsmodels-2007","downloads":0,"creationDate":"2017-09-14T16:34:36.938Z","title":"Load estimation and control using learned dynamics models","author_short":["Petkos, G.","Vijayakumar, S."],"year":2007,"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/jfslin/jfslin.github.io/master/jf2lin.bib","bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"Load estimation and control using learned dynamics models","author":[{"propositions":[],"lastnames":["Petkos"],"firstnames":["G."],"suffixes":[]},{"propositions":[],"lastnames":["Vijayakumar"],"firstnames":["S."],"suffixes":[]}],"booktitle":"Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems","year":"2007","pages":"1527--1532","abstract":"Classic adaptive control methods for handling varying <span class='snippet'>loads</span> rely on an analytically derived model of the robot's dynamics. However, in many situations, it is not feasible or easy to obtain an accurate analytic model of the robot's dynamics. An alternative to analytically deriving the dynamics is learning the dynamics from movement data. This paper describes a <span class='snippet'>load</span> <span class='snippet'>estimation</span> technique that uses the learned instead of analytically derived dynamics. We study examples where the various inertial parameters of the <span class='snippet'>load</span> are estimated from the learned models, their effectiveness in control is evaluated along with their robustness in light of imperfect, intermediate dynamic models.","doi":"10.1109/IROS.2007.4399373","review":"- How can we use adaptive control to pick up loads? Picking up things changes dynamics of the system. Two ways of this happening: - q, dq, ddq sensing: special case of parameter estimation. The whole 10 parameters thing. - however, not all parameters can be identified - some params are not excited, while others are linearly dependent - could get equations analytically, but often complicated - so we'll numerically learn it instead - another way to do this is to look at torque sensing and map input torque to model - need to use a PD controller, then switched a \"composite\" controller (PD + model) - so the paper combines these to figure out load weight - so convert a nx10*n param system to a 11 model system (10 for the last link, 1 for the rest of the robot) - estimates model accuracy by looking at the ratio between fdfwd (which the model generates) term and fdbk load","timestamp":"2011.02.03","bibtex":"@InProceedings{Petkos2007,\n Title = {Load estimation and control using learned dynamics models},\n Author = {Petkos, G. and Vijayakumar, S.},\n Booktitle = {Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems},\n Year = {2007},\n Pages = {1527--1532},\n\n Abstract = {Classic adaptive control methods for handling varying <span class='snippet'>loads</span> rely on an analytically derived model of the robot's dynamics. However, in many situations, it is not feasible or easy to obtain an accurate analytic model of the robot's dynamics. An alternative to analytically deriving the dynamics is learning the dynamics from movement data. This paper describes a <span class='snippet'>load</span> <span class='snippet'>estimation</span> technique that uses the learned instead of analytically derived dynamics. We study examples where the various inertial parameters of the <span class='snippet'>load</span> are estimated from the learned models, their effectiveness in control is evaluated along with their robustness in light of imperfect, intermediate dynamic models.},\n Doi = {10.1109/IROS.2007.4399373},\n Review = {- How can we use adaptive control to pick up loads? Picking up things changes dynamics of the system. Two ways of this happening:\n- q, dq, ddq sensing: special case of parameter estimation. The whole 10 parameters thing. \n - however, not all parameters can be identified\n - some params are not excited, while others are linearly dependent\n - could get equations analytically, but often complicated\n - so we'll numerically learn it instead\n- another way to do this is to look at torque sensing and map input torque to model\n - need to use a PD controller, then switched a \"composite\" controller (PD + model)\n\n- so the paper combines these to figure out load weight\n - so convert a nx10*n param system to a 11 model system (10 for the last link, 1 for the rest of the robot)\n- estimates model accuracy by looking at the ratio between fdfwd (which the model generates) term and fdbk load},\n Timestamp = {2011.02.03}\n}\n\n","author_short":["Petkos, G.","Vijayakumar, S."],"key":"Petkos2007","id":"Petkos2007","bibbaseid":"petkos-vijayakumar-loadestimationandcontrolusinglearneddynamicsmodels-2007","role":"author","urls":{},"downloads":0},"search_terms":["load","estimation","control","using","learned","dynamics","models","petkos","vijayakumar"],"keywords":[],"authorIDs":[],"dataSources":["iCsmKnycRmHPxmhBd"]}