The paper introduces an approach to generate control strategies for cable handling tasks automatically, using Reinforcement Learning. Cable handling tasks require human like sensomotoric and cognitive abilities. Humans develop an understanding of the cable behaviour and are able to derive control strategies, how to handle the cable. Transferring these strategies to an industrial robot control system with the usual programming interface is very difficult. The introduced approach addresses this problem by giving the robot control the possibility to learn the cable handling strategies by itself. Therefore an Virtual Commissioning simulation environment was adapted to model Reinforcement Learning problems using the OpenAI Gym and Functional Mock-Up Interface standard.
%0 Conference Paper
%1 9253103
%A Jaensch, Florian
%A Verl, Alexander
%B 2020 Third International Conference on Artificial Intelligence for Industries (AI4I)
%D 2020
%K isw myown simulation
%P 27-31
%R 10.1109/AI4I49448.2020.00013
%T Virtual Commissioning Simulation as Reinforcement Learning Environment for Robot Cable Handling
%X The paper introduces an approach to generate control strategies for cable handling tasks automatically, using Reinforcement Learning. Cable handling tasks require human like sensomotoric and cognitive abilities. Humans develop an understanding of the cable behaviour and are able to derive control strategies, how to handle the cable. Transferring these strategies to an industrial robot control system with the usual programming interface is very difficult. The introduced approach addresses this problem by giving the robot control the possibility to learn the cable handling strategies by itself. Therefore an Virtual Commissioning simulation environment was adapted to model Reinforcement Learning problems using the OpenAI Gym and Functional Mock-Up Interface standard.
@inproceedings{9253103,
abstract = {The paper introduces an approach to generate control strategies for cable handling tasks automatically, using Reinforcement Learning. Cable handling tasks require human like sensomotoric and cognitive abilities. Humans develop an understanding of the cable behaviour and are able to derive control strategies, how to handle the cable. Transferring these strategies to an industrial robot control system with the usual programming interface is very difficult. The introduced approach addresses this problem by giving the robot control the possibility to learn the cable handling strategies by itself. Therefore an Virtual Commissioning simulation environment was adapted to model Reinforcement Learning problems using the OpenAI Gym and Functional Mock-Up Interface standard.},
added-at = {2022-03-23T16:37:47.000+0100},
author = {Jaensch, Florian and Verl, Alexander},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/241ab4eaea79e03cbcf8894da460bc24a/isw-bibliothek},
booktitle = {2020 Third International Conference on Artificial Intelligence for Industries (AI4I)},
doi = {10.1109/AI4I49448.2020.00013},
interhash = {f043f8169c9681adb6cfb63449f15263},
intrahash = {41ab4eaea79e03cbcf8894da460bc24a},
keywords = {isw myown simulation},
month = {9},
pages = {27-31},
timestamp = {2022-03-23T15:38:14.000+0100},
title = {Virtual Commissioning Simulation as Reinforcement Learning Environment for Robot Cable Handling},
year = 2020
}
