Generalizing Behavior Trees and Motion-Generator (BTMG) Policy Representation for Robotic Tasks Over Scenario Parameters

Faseeh Ahmad; Matthias Mayr; Elin Anna Topp; Jacek Malec; Volker Krueger

Generalizing Behavior Trees and Motion-Generator (BTMG) Policy Representation for Robotic Tasks Over Scenario Parameters

Faseeh Ahmad, Matthias Mayr, Elin Anna Topp, Jacek Malec, Volker Krueger

Forskningsoutput: Konferensbidrag › Konferenspaper, ej i proceeding/ej förlagsutgivet › Peer review

Sammanfattning

We propose a generalisation of a behaviour tree and motiongenerator based robot arm policy representation for learning and solving tasks such as contact-rich tasks like peg insertion or pushing an object. We use planning to generate skill sequences needed to execute these tasks and rely on reinforcement learning to obtain parameters of the policy. We assume gaussian processes as a suitable method for this generalisation and present preliminary, promising results from initial experiments.

Originalspråk	engelska
Antal sidor	2
Status	Published - 2022

Ämnesklassifikation (UKÄ)

Robotik och automation

Tillgång till dokument

https://prl-theworkshop.github.io/prl2022-ijcai/papers/PRL2022_paper_25.pdfLicens: Ospecificerad

2 Konferenspaper i proceeding
1 Artikel i vetenskaplig tidskrift
1 Doktorsavhandling (sammanläggning)

A C++ Implementation of a Cartesian Impedance Controller for Robotic Manipulators
Mayr, M. & Salt Ducaju, J., 2024 jan. 23, I: Journal of Open Source Software. 9, 93, s. 5194-5200 6 s.
Forskningsoutput: Tidskriftsbidrag › Artikel i vetenskaplig tidskrift › Peer review

Öppen tillgång
Learning with Skill-based Robot Systems: Combining Planning & Knowledge Representation with Reinforcement Learning
Mayr, M., 2024 jan. 9, Lund: Computer Science, Lund University. 286 s.
Forskningsoutput: Avhandling › Doktorsavhandling (sammanläggning)

Öppen tillgång
Fil
SkiROS2: A Skill-Based Robot Control Platform for ROS
Mayr, M., Rovida, F. & Krueger, V., 2023 dec. 13, 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE - Institute of Electrical and Electronics Engineers Inc., s. 6273-6280 8 s.
Forskningsoutput: Kapitel i bok/rapport/Conference proceeding › Konferenspaper i proceeding › Peer review

1 Avslutade
1 Ej startade
2 Aktiva

RSS: Robotics and Semantic Systems
Krueger, V. (Forskare), Topp, E. A. (Forskare), Malec, J. (Forskare), Haage, M. (Forskare), Mayr, M. (Forskare), Åström, H. (Forskare), Dürr, A. (Forskare), Nugues, P. (Forskare), Klang, M. (Forskare), Klöckner, M. (Forskare), Lager, M. (Forskare), Robertz, S. (Forskare), Nardi, L. (Forskare) & Nilsson, K. (Forskare)
2010/01/01 → …
Projekt: Nätverk
RobotLab LTH
Bagge Carlson, F. (Forskare), Johansson, R. (Forskare), Karlsson, M. (Forskare), Olofsson, B. (Forskare), Robertsson, A. (Forskare), Robertz, S. (Forskare), Haage, M. (Forskare), Malec, J. (Forskare), Nilsson, K. (Forskare), Nugues, P. (Forskare), Stenmark, M. (Forskare), Topp, E. A. (Forskare), Krueger, V. (Forskare), Åström, H. (Forskare), Mayr, M. (Forskare), Salt Ducaju, J. (Forskare), Nishimura, M. (Administratör), Wisbrant, J. (Projektkommunikatör), Dürr, A. (Forskare), Mayr, M. (Forskare), Nugues, P. (Forskare), Klang, M. (Forskningsingenjör), Klöckner, M. (Forskare), Nardi, L. (Forskare), Ahmad, F. (Forskare), Oxenstierna, J. (Forskare), Rizwan, M. (Forskare), Reichenbach, C. (Forskare), Bergström, J. (Forskare), Dell'Unto, N. (Forskare), Maunsbach, L. (Forskare), Åström, K. (Forskare), Blomdell, A. (Forskningsingenjör), Magnusson, M. (Forskare), Fransson, P.-A. (Forskare), Karayiannidis, Y. (Forskare), Johansson, A. T. (Forskare), Jia, Z. (Forskare), Laban, L. (Forskare), Wingqvist, B. (Forskare), Guberina, M. (Forskare), Jena, A. (Forskare), Westin, E. (Administratör), Frick, C. (Administratör), Pisarevskiy, A. (Forskningsingenjör), Nilsson, A. (Forskningsingenjör), Reitmann, S. (Forskare), Hvarfner, C. (Forskare), Stoltenberg, P. (Forskare) & Fregnan, S. (Forskare)
1993/01/01 → …
Projekt: Forskning
WASP Professor Package: Cognitive Robots for Manufacturing
Krueger, V. (Forskare), Ahmad, F. (Forskarstuderande), Mayr, M. (Forskarstuderande), Oxenstierna, J. (Forskarstuderande), Åström, H. (Forskarstuderande) & Wisbrant, J. (Forskare)
Projekt: Forskning

RobotLab LTH Infrastruktur
Krueger, V. (Manager), Olofsson, B. (Manager), Karayiannidis, Y. (Manager), Haage, M. (Manager), Malec, J. (Manager) & Topp, E. A. (Manager)
Lunds Tekniska Högskola
Infrastruktur

Citera det här

@conference{aed2dfb2034346559f2529852c81eb70,

title = "Generalizing Behavior Trees and Motion-Generator (BTMG) Policy Representation for Robotic Tasks Over Scenario Parameters",

abstract = "We propose a generalisation of a behaviour tree and motiongenerator based robot arm policy representation for learning and solving tasks such as contact-rich tasks like peg insertion or pushing an object. We use planning to generate skill sequences needed to execute these tasks and rely on reinforcement learning to obtain parameters of the policy. We assume gaussian processes as a suitable method for this generalisation and present preliminary, promising results from initial experiments.",

keywords = "Generalization, Industrial Robots, Reinforcement learning",

author = "Faseeh Ahmad and Matthias Mayr and Topp, {Elin Anna} and Jacek Malec and Volker Krueger",

year = "2022",

language = "English",

}

TY - CONF

T1 - Generalizing Behavior Trees and Motion-Generator (BTMG) Policy Representation for Robotic Tasks Over Scenario Parameters

AU - Ahmad, Faseeh

AU - Mayr, Matthias

AU - Topp, Elin Anna

AU - Malec, Jacek

AU - Krueger, Volker

PY - 2022

Y1 - 2022

N2 - We propose a generalisation of a behaviour tree and motiongenerator based robot arm policy representation for learning and solving tasks such as contact-rich tasks like peg insertion or pushing an object. We use planning to generate skill sequences needed to execute these tasks and rely on reinforcement learning to obtain parameters of the policy. We assume gaussian processes as a suitable method for this generalisation and present preliminary, promising results from initial experiments.

AB - We propose a generalisation of a behaviour tree and motiongenerator based robot arm policy representation for learning and solving tasks such as contact-rich tasks like peg insertion or pushing an object. We use planning to generate skill sequences needed to execute these tasks and rely on reinforcement learning to obtain parameters of the policy. We assume gaussian processes as a suitable method for this generalisation and present preliminary, promising results from initial experiments.

KW - Generalization

KW - Industrial Robots

KW - Reinforcement learning

M3 - Paper, not in proceeding

ER -

Generalizing Behavior Trees and Motion-Generator (BTMG) Policy Representation for Robotic Tasks Over Scenario Parameters

Sammanfattning

Ämnesklassifikation (UKÄ)

Tillgång till dokument

Fingeravtryck

Forskningsoutput

A C++ Implementation of a Cartesian Impedance Controller for Robotic Manipulators

Learning with Skill-based Robot Systems: Combining Planning & Knowledge Representation with Reinforcement Learning

SkiROS2: A Skill-Based Robot Control Platform for ROS

Projekt

RSS: Robotics and Semantic Systems

RobotLab LTH

WASP Professor Package: Cognitive Robots for Manufacturing

Utrustning

RobotLab LTH Infrastruktur

Citera det här