PyRoboLearn: A Python Framework for Robot Learning Practitioners

Brian Delhaisse, Leonel Rozo, Darwin G. Caldwell
; Proceedings of the Conference on Robot Learning, PMLR 100:1348-1358, 2020.

Abstract

On the quest for building autonomous robots, several robot learning frameworks with different functionalities have recently been developed. Yet, frameworks that combine diverse learning paradigms (such as imitation and reinforcement learning) into a common place are scarce. Existing ones tend to be robot-specific, and often require time-consuming work to be used with other robots. Also, their architecture is often weakly structured, mainly because of a lack of modularity and flexibility. This leads users to reimplement several pieces of code to integrate them into their own experimental or benchmarking work. To overcome these issues, we introduce PyRoboLearn, a new Python robot learning framework that combines different learning paradigms into a single framework. Our framework provides a plethora of robotic environments, learning models and algorithms. PyRoboLearn is developed with a particular focus on modularity, flexibility, generality, and simplicity to favor (re)usability. This is achieved by abstracting each key concept, undertaking a modular programming approach, minimizing the coupling among the different modules, and favoring composition over inheritance for better flexibility. We demonstrate the different features and utility of our framework through different use cases.

Cite this Paper


BibTeX
@InProceedings{pmlr-v100-delhaisse20a, title = {PyRoboLearn: A Python Framework for Robot Learning Practitioners}, author = {Delhaisse, Brian and Rozo, Leonel and Caldwell, Darwin G.}, booktitle = {Proceedings of the Conference on Robot Learning}, pages = {1348--1358}, year = {2020}, editor = {Leslie Pack Kaelbling and Danica Kragic and Komei Sugiura}, volume = {100}, series = {Proceedings of Machine Learning Research}, address = {}, month = {30 Oct--01 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v100/delhaisse20a/delhaisse20a.pdf}, url = {http://proceedings.mlr.press/v100/delhaisse20a.html}, abstract = {On the quest for building autonomous robots, several robot learning frameworks with different functionalities have recently been developed. Yet, frameworks that combine diverse learning paradigms (such as imitation and reinforcement learning) into a common place are scarce. Existing ones tend to be robot-specific, and often require time-consuming work to be used with other robots. Also, their architecture is often weakly structured, mainly because of a lack of modularity and flexibility. This leads users to reimplement several pieces of code to integrate them into their own experimental or benchmarking work. To overcome these issues, we introduce PyRoboLearn, a new Python robot learning framework that combines different learning paradigms into a single framework. Our framework provides a plethora of robotic environments, learning models and algorithms. PyRoboLearn is developed with a particular focus on modularity, flexibility, generality, and simplicity to favor (re)usability. This is achieved by abstracting each key concept, undertaking a modular programming approach, minimizing the coupling among the different modules, and favoring composition over inheritance for better flexibility. We demonstrate the different features and utility of our framework through different use cases.} }
Endnote
%0 Conference Paper %T PyRoboLearn: A Python Framework for Robot Learning Practitioners %A Brian Delhaisse %A Leonel Rozo %A Darwin G. Caldwell %B Proceedings of the Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2020 %E Leslie Pack Kaelbling %E Danica Kragic %E Komei Sugiura %F pmlr-v100-delhaisse20a %I PMLR %J Proceedings of Machine Learning Research %P 1348--1358 %U http://proceedings.mlr.press %V 100 %W PMLR %X On the quest for building autonomous robots, several robot learning frameworks with different functionalities have recently been developed. Yet, frameworks that combine diverse learning paradigms (such as imitation and reinforcement learning) into a common place are scarce. Existing ones tend to be robot-specific, and often require time-consuming work to be used with other robots. Also, their architecture is often weakly structured, mainly because of a lack of modularity and flexibility. This leads users to reimplement several pieces of code to integrate them into their own experimental or benchmarking work. To overcome these issues, we introduce PyRoboLearn, a new Python robot learning framework that combines different learning paradigms into a single framework. Our framework provides a plethora of robotic environments, learning models and algorithms. PyRoboLearn is developed with a particular focus on modularity, flexibility, generality, and simplicity to favor (re)usability. This is achieved by abstracting each key concept, undertaking a modular programming approach, minimizing the coupling among the different modules, and favoring composition over inheritance for better flexibility. We demonstrate the different features and utility of our framework through different use cases.
APA
Delhaisse, B., Rozo, L. & Caldwell, D.G.. (2020). PyRoboLearn: A Python Framework for Robot Learning Practitioners. Proceedings of the Conference on Robot Learning, in PMLR 100:1348-1358

Related Material